DFHack Lua API

DFHack has extensive support for the Lua scripting language, providing access to:

  1. Raw data structures used by the game.
  2. Many C++ functions for high-level access to these structures, and interaction with dfhack itself.
  3. Some functions exported by C++ plugins.

Lua code can be used both for writing scripts, which are treated by DFHack command line prompt almost as native C++ commands, and invoked by plugins written in C++.

This document describes native API available to Lua in detail. It does not describe all of the utility functions implemented by Lua files located in hack/lua/* (library/lua/* in the git repo).

DF data structure wrapper

Data structures of the game are defined in XML files located in library/xml (and online, and automatically exported to lua code as a tree of objects and functions under the df global, which also broadly maps to the df namespace in the headers generated for C++.


The wrapper provides almost raw access to the memory of the game, so mistakes in manipulating objects are as likely to crash the game as equivalent plain C++ code would be - e.g. null pointer access is safely detected, but dangling pointers aren’t.

Objects managed by the wrapper can be broadly classified into the following groups:

  1. Typed object pointers (references).

    References represent objects in DF memory with a known type.

    In addition to fields and methods defined by the wrapped type, every reference has some built-in properties and methods.

  2. Untyped pointers

    Represented as lightuserdata.

    In assignment to a pointer NULL can be represented either as nil, or a NULL lightuserdata; reading a NULL pointer field returns nil.

  3. Named types

    Objects in the df tree that represent identity of struct, class, enum and bitfield types. They host nested named types, static methods, builtin properties & methods, and, for enums and bitfields, the bi-directional mapping between key names and values.

  4. The global object

    df.global corresponds to the df::global namespace, and behaves as a mix between a named type and a reference, containing both nested types and fields corresponding to global symbols.

In addition to the global object and top-level types the df global also contains a few global builtin utility functions.

Typed object references

The underlying primitive lua object is userdata with a metatable. Every structured field access produces a new userdata instance.

All typed objects have the following built-in features:

  • ref1 == ref2, tostring(ref)

    References implement equality by type & pointer value, and string conversion.

  • pairs(ref)

    Returns an iterator for the sequence of actual C++ field names and values. Fields are enumerated in memory order. Methods and lua wrapper properties are not included in the iteration.


    a few of the data structures (like ui_look_list) contain unions with pointers to different types with vtables. Using pairs on such structs is an almost sure way to crash with an access violation.

  • ref._kind

    Returns one of: primitive, struct, container, or bitfield, as appropriate for the referenced object.

  • ref._type

    Returns the named type object or a string that represents the referenced object type.

  • ref:sizeof()

    Returns size, address

  • ref:new()

    Allocates a new instance of the same type, and copies data from the current object.

  • ref:delete()

    Destroys the object with the C++ delete operator. If the destructor is not available, returns false. (This typically only occurs when trying to delete an instance of a DF class with virtual methods whose vtable address has not been found; it is impossible for delete() to determine the validity of ref.)


    ref must be an object allocated with new, like in C++. Calling obj.field:delete() where obj was allocated with new will not work. After delete() returns, ref remains as a dangling pointer, like a raw C++ pointer would. Any accesses to ref after ref:delete() has been called are undefined behavior.

  • ref:assign(object)

    Assigns data from object to ref. Object must either be another ref of a compatible type, or a lua table; in the latter case special recursive assignment rules are applied.

  • ref:_displace(index[,step])

    Returns a new reference with the pointer adjusted by index*step. Step defaults to the natural object size.

Primitive references

References of the _kind 'primitive' are used for objects that don’t fit any of the other reference types. Such references can only appear as a value of a pointer field, or as a result of calling the _field() method.

They behave as structs with a value field of the right type. If the object’s XML definition has a ref-target attribute, they will also have a read-only ref_target field set to the corresponding type object.

To make working with numeric buffers easier, they also allow numeric indices. Note that other than excluding negative values no bound checking is performed, since buffer length is not available. Index 0 is equivalent to the value field.

Struct references

Struct references are used for class and struct objects.

They implement the following features:

  • ref.field, ref.field = value

    Valid fields of the structure may be accessed by subscript.

    Primitive typed fields, i.e. numbers & strings, are converted to/from matching lua values. The value of a pointer is a reference to the target, or nil/NULL. Complex types are represented by a reference to the field within the structure; unless recursive lua table assignment is used, such fields can only be read.


    In case of inheritance, superclass fields have precedence over the subclass, but fields shadowed in this way can still be accessed as ref['subclasstype.field'].

    This shadowing order is necessary because vtable-based classes are automatically exposed in their exact type, and the reverse rule would make access to superclass fields unreliable.

  • ref._field(field)

    Returns a reference to a valid field. That is, unlike regular subscript, it returns a reference to the field within the structure even for primitive typed fields and pointers.

  • ref:vmethod(args...)

    Named virtual methods are also exposed, subject to the same shadowing rules.

  • pairs(ref)

    Enumerates all real fields (but not methods) in memory order, which is the same as declaration order.

Container references

Containers represent vectors and arrays, possibly resizable.

A container field can associate an enum to the container reference, which allows accessing elements using string keys instead of numerical indices.

Note that two-dimensional arrays in C++ (ie pointers to pointers) are exposed to lua as one-dimensional. The best way to handle this is probably array[x].value:_displace(y).

Implemented features:

  • ref._enum

    If the container has an associated enum, returns the matching named type object.

  • #ref

    Returns the length of the container.

  • ref[index]

    Accesses the container element, using either a 0-based numerical index, or, if an enum is associated, a valid enum key string.

    Accessing an invalid index is an error, but some container types may return a default value, or auto-resize instead for convenience. Currently this relaxed mode is implemented by df-flagarray aka BitArray.

  • ref._field(index)

    Like with structs, returns a pointer to the array element, if possible. Flag and bit arrays cannot return such pointer, so it fails with an error.

  • pairs(ref), ipairs(ref)

    If the container has no associated enum, both behave identically, iterating over numerical indices in order. Otherwise, ipairs still uses numbers, while pairs tries to substitute enum keys whenever possible.

  • ref:resize(new_size)

    Resizes the container if supported, or fails with an error.

  • ref:insert(index,item)

    Inserts a new item at the specified index. To add at the end, use #ref, or just '#' as index.

  • ref:erase(index)

    Removes the element at the given valid index.

Bitfield references

Bitfields behave like special fixed-size containers. Consider them to be something in between structs and fixed-size vectors.

The _enum property points to the bitfield type. Numerical indices correspond to the shift value, and if a subfield occupies multiple bits, the ipairs order would have a gap.

Since currently there is no API to allocate a bitfield object fully in GC-managed lua heap, consider using the lua table assignment feature outlined below in order to pass bitfield values to dfhack API functions that need them, e.g. matinfo:matches{metal=true}.

Named types

Named types are exposed in the df tree with names identical to the C++ version, except for the :: vs . difference.

All types and the global object have the following features:

  • type._kind

    Evaluates to one of struct-type, class-type, enum-type, bitfield-type or global.

  • type._identity

    Contains a lightuserdata pointing to the underlying DFHack::type_instance object.

Types excluding the global object also support:

  • type:sizeof()

    Returns the size of an object of the type.

  • type:new()

    Creates a new instance of an object of the type.

  • type:is_instance(object)

    Returns true if object is same or subclass type, or a reference to an object of same or subclass type. It is permissible to pass nil, NULL or non-wrapper value as object; in this case the method returns nil.

In addition to this, enum and bitfield types contain a bi-directional mapping between key strings and values, and also map _first_item and _last_item to the min and max values.

Struct and class types with instance-vector attribute in the xml have a type.find(key) function that wraps the find method provided in C++.

Global functions

The df table itself contains the following functions and values:

  • NULL, df.NULL

    Contains the NULL lightuserdata.

  • df.isnull(obj)

    Evaluates to true if obj is nil or NULL; false otherwise.

  • df.isvalid(obj[,allow_null])

    For supported objects returns one of type, voidptr, ref.

    If allow_null is true, and obj is nil or NULL, returns null.

    Otherwise returns nil.

  • df.sizeof(obj)

    For types and refs identical to obj:sizeof(). For lightuserdata returns nil, address

  • df.new(obj), df.delete(obj), df.assign(obj, obj2)

    Equivalent to using the matching methods of obj.

  • df._displace(obj,index[,step])

    For refs equivalent to the method, but also works with lightuserdata (step is mandatory then).

  • df.is_instance(type,obj)

    Equivalent to the method, but also allows a reference as proxy for its type.

  • df.new(ptype[,count])

    Allocate a new instance, or an array of built-in types. The ptype argument is a string from the following list: string, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t, int64_t, uint64_t, bool, float, double. All of these except string can be used with the count argument to allocate an array.

  • df.reinterpret_cast(type,ptr)

    Converts ptr to a ref of specified type. The type may be anything acceptable to df.is_instance. Ptr may be nil, a ref, a lightuserdata, or a number.

    Returns nil if NULL, or a ref.

Recursive table assignment

Recursive assignment is invoked when a lua table is assigned to a C++ object or field, i.e. one of:

  • ref:assign{...}
  • ref.field = {...}

The general mode of operation is that all fields of the table are assigned to the fields of the target structure, roughly emulating the following code:

function rec_assign(ref,table)
    for key,value in pairs(table) do
        ref[key] = value

Since assigning a table to a field using = invokes the same process, it is recursive.

There are however some variations to this process depending on the type of the field being assigned to:

  1. If the table contains an assign field, it is applied first, using the ref:assign(value) method. It is never assigned as a usual field.

  2. When a table is assigned to a non-NULL pointer field using the ref.field = {...} syntax, it is applied to the target of the pointer instead.

    If the pointer is NULL, the table is checked for a new field:

    1. If it is nil or false, assignment fails with an error.
    2. If it is true, the pointer is initialized with a newly allocated object of the declared target type of the pointer.
    3. Otherwise, table.new must be a named type, or an object of a type compatible with the pointer. The pointer is initialized with the result of calling table.new:new().

    After this auto-vivification process, assignment proceeds as if the pointer wasn’t NULL.

    Obviously, the new field inside the table is always skipped during the actual per-field assignment processing.

  3. If the target of the assignment is a container, a separate rule set is used:

    1. If the table contains neither assign nor resize fields, it is interpreted as an ordinary 1-based lua array. The container is resized to the #-size of the table, and elements are assigned in numeric order:

      for i=1,#table do ref[i-1] = table[i] end
    2. Otherwise, resize must be true, false, or an explicit number. If it is not false, the container is resized. After that the usual struct-like ‘pairs’ assignment is performed.

      In case resize is true, the size is computed by scanning the table for the largest numeric key.

    This means that in order to reassign only one element of a container using this system, it is necessary to use:

    { resize=false, [idx]=value }

Since nil inside a table is indistinguishable from missing key, it is necessary to use df.NULL as a null pointer value.

This system is intended as a way to define a nested object tree using pure lua data structures, and then materialize it in C++ memory in one go. Note that if pointer auto-vivification is used, an error in the middle of the recursive walk would not destroy any objects allocated in this way, so the user should be prepared to catch the error and do the necessary cleanup.


DFHack utility functions are placed in the dfhack global tree.

Native utilities

Input & Output

  • dfhack.print(args...)

    Output tab-separated args as standard lua print would do, but without a newline.

  • print(args...), dfhack.println(args...)

    A replacement of the standard library print function that works with DFHack output infrastructure.

  • dfhack.printerr(args...)

    Same as println; intended for errors. Uses red color and logs to stderr.log.

  • dfhack.color([color])

    Sets the current output color. If color is nil or -1, resets to default. Returns the previous color value.

  • dfhack.is_interactive()

    Checks if the thread can access the interactive console and returns true or false.

  • dfhack.lineedit([prompt[,history_filename]])

    If the thread owns the interactive console, shows a prompt and returns the entered string. Otherwise returns nil, error.

    Depending on the context, this function may actually yield the running coroutine and let the C++ code release the core suspend lock. Using an explicit dfhack.with_suspend will prevent this, forcing the function to block on input with lock held.

  • dfhack.interpreter([prompt[,history_filename[,env]]])

    Starts an interactive lua interpreter, using the specified prompt string, global environment and command-line history file.

    If the interactive console is not accessible, returns nil, error.

Exception handling

  • dfhack.error(msg[,level[,verbose]])

    Throws a dfhack exception object with location and stack trace. The verbose parameter controls whether the trace is printed by default.

  • qerror(msg[,level])

    Calls dfhack.error() with verbose being false. Intended to be used for user-caused errors in scripts, where stack traces are not desirable.

  • dfhack.pcall(f[,args...])

    Invokes f via xpcall, using an error function that attaches a stack trace to the error. The same function is used by SafeCall in C++, and dfhack.safecall.

  • safecall(f[,args...]), dfhack.safecall(f[,args...])

    Just like pcall, but also prints the error using printerr before returning. Intended as a convenience function.

  • dfhack.saferesume(coroutine[,args...])

    Compares to coroutine.resume like dfhack.safecall vs pcall.

  • dfhack.exception

    Metatable of error objects used by dfhack. The objects have the following properties:


    The location prefix string, or nil.


    The base message string.


    The stack trace string, or nil.


    A different exception object, or nil.


    The coroutine that has thrown the exception.


    Boolean, or nil; specifies if where and stacktrace should be printed.

    tostring(err), or err:tostring([verbose])

    Converts the exception to string.

  • dfhack.exception.verbose

    The default value of the verbose argument of err:tostring().


  • dfhack.VERSION

    DFHack version string constant.

  • dfhack.curry(func,args...), or curry(func,args...)

    Returns a closure that invokes the function with args combined both from the curry call and the closure call itself. I.e. curry(func,a,b)(c,d) equals func(a,b,c,d).

Locking and finalization

  • dfhack.with_suspend(f[,args...])

    Calls f with arguments after grabbing the DF core suspend lock. Suspending is necessary for accessing a consistent state of DF memory.

    Returned values and errors are propagated through after releasing the lock. It is safe to nest suspends.

    Every thread is allowed only one suspend per DF frame, so it is best to group operations together in one big critical section. A plugin can choose to run all lua code inside a C++-side suspend lock.

  • dfhack.call_with_finalizer(num_cleanup_args,always,cleanup_fn[,cleanup_args...],fn[,args...])

    Invokes fn with args, and after it returns or throws an error calls cleanup_fn with cleanup_args. Any return values from fn are propagated, and errors are re-thrown.

    The num_cleanup_args integer specifies the number of cleanup_args, and the always boolean specifies if cleanup should be called in any case, or only in case of an error.

  • dfhack.with_finalize(cleanup_fn,fn[,args...])

    Calls fn with arguments, then finalizes with cleanup_fn. Implemented using call_with_finalizer(0,true,...).

  • dfhack.with_onerror(cleanup_fn,fn[,args...])

    Calls fn with arguments, then finalizes with cleanup_fn on any thrown error. Implemented using call_with_finalizer(0,false,...).

  • dfhack.with_temp_object(obj,fn[,args...])

    Calls fn(obj,args...), then finalizes with obj:delete().

Persistent configuration storage

This api is intended for storing configuration options in the world itself. It probably should be restricted to data that is world-dependent.

Entries are identified by a string key, but it is also possible to manage multiple entries with the same key; their identity is determined by entry_id. Every entry has a mutable string value, and an array of 7 mutable ints.

  • dfhack.persistent.get(key), entry:get()

    Retrieves a persistent config record with the given string key, or refreshes an already retrieved entry. If there are multiple entries with the same key, it is undefined which one is retrieved by the first version of the call.

    Returns entry, or nil if not found.

  • dfhack.persistent.delete(key), entry:delete()

    Removes an existing entry. Returns true if succeeded.

  • dfhack.persistent.get_all(key[,match_prefix])

    Retrieves all entries with the same key, or starting with key..’/’. Calling get_all('',true) will match all entries.

    If none found, returns nil; otherwise returns an array of entries.

  • dfhack.persistent.save({key=str1, ...}[,new]), entry:save([new])

    Saves changes in an entry, or creates a new one. Passing true as new forces creation of a new entry even if one already exists; otherwise the existing one is simply updated. Returns entry, did_create_new

Since the data is hidden in data structures owned by the DF world, and automatically stored in the save game, these save and retrieval functions can just copy values in memory without doing any actual I/O. However, currently every entry has a 180+-byte dead-weight overhead.

It is also possible to associate one bit per map tile with an entry, using these two methods:

  • entry:getTilemask(block[, create])

    Retrieves the tile bitmask associated with this entry in the given map block. If create is true, an empty mask is created if none exists; otherwise the function returns nil, which must be assumed to be the same as an all-zero mask.

  • entry:deleteTilemask(block)

    Deletes the associated tile mask from the given map block.

Note that these masks are only saved in fortress mode, and also that deleting the persistent entry will NOT delete the associated masks.

Material info lookup

A material info record has fields:

  • type, index, material

    DF material code pair, and a reference to the material object.

  • mode

    One of 'builtin', 'inorganic', 'plant', 'creature'.

  • inorganic, plant, creature

    If the material is of the matching type, contains a reference to the raw object.

  • figure

    For a specific creature material contains a ref to the historical figure.


  • dfhack.matinfo.decode(type,index)

    Looks up material info for the given number pair; if not found, returs nil.

  • ....decode(matinfo), ....decode(item), ....decode(obj)

    Uses matinfo.type/matinfo.index, item getter vmethods, or obj.mat_type/obj.mat_index to get the code pair.

  • dfhack.matinfo.find(token[,token...])

    Looks up material by a token string, or a pre-split string token sequence.

  • dfhack.matinfo.getToken(...), info:getToken()

    Applies decode and constructs a string token.

  • info:toString([temperature[,named]])

    Returns the human-readable name at the given temperature.

  • info:getCraftClass()

    Returns the classification used for craft skills.

  • info:matches(obj)

    Checks if the material matches job_material_category or job_item. Accept dfhack_material_category auto-assign table.

Random number generation

  • dfhack.random.new([seed[,perturb_count]])

    Creates a new random number generator object. Without any arguments, the object is initialized using current time. Otherwise, the seed must be either a non-negative integer, or a list of such integers. The second argument may specify the number of additional randomization steps performed to improve the initial state.

  • rng:init([seed[,perturb_count]])

    Re-initializes an already existing random number generator object.

  • rng:random([limit])

    Returns a random integer. If limit is specified, the value is in the range [0, limit); otherwise it uses the whole 32-bit unsigned integer range.

  • rng:drandom()

    Returns a random floating-point number in the range [0,1).

  • rng:drandom0()

    Returns a random floating-point number in the range (0,1).

  • rng:drandom1()

    Returns a random floating-point number in the range [0,1].

  • rng:unitrandom()

    Returns a random floating-point number in the range [-1,1].

  • rng:unitvector([size])

    Returns multiple values that form a random vector of length 1, uniformly distributed over the corresponding sphere surface. The default size is 3.

  • fn = rng:perlin([dim]); fn(x[,y[,z]])

    Returns a closure that computes a classical Perlin noise function of dimension dim, initialized from this random generator. Dimension may be 1, 2 or 3 (default).

C++ function wrappers

Thin wrappers around C++ functions, similar to the ones for virtual methods. One notable difference is that these explicit wrappers allow argument count adjustment according to the usual lua rules, so trailing false/nil arguments can be omitted.

  • dfhack.getOSType()

    Returns the OS type string from symbols.xml.

  • dfhack.getDFVersion()

    Returns the DF version string from symbols.xml.

  • dfhack.getDFHackVersion()

  • dfhack.getDFHackRelease()

  • dfhack.getDFHackBuildID()

  • dfhack.getCompiledDFVersion()

  • dfhack.getGitDescription()

  • dfhack.getGitCommit()

  • dfhack.getGitXmlCommit()

  • dfhack.getGitXmlExpectedCommit()

  • dfhack.gitXmlMatch()

  • dfhack.isRelease()

    Return information about the DFHack build in use.


    getCompiledDFVersion() returns the DF version specified at compile time, while getDFVersion() returns the version and typically the OS as well. These do not necessarily match - for example, DFHack 0.34.11-r5 worked with DF 0.34.10 and 0.34.11, so the former function would always return 0.34.11 while the latter would return v0.34.10 <platform> or v0.34.11 <platform>.

  • dfhack.getDFPath()

    Returns the DF directory path.

  • dfhack.getHackPath()

    Returns the dfhack directory path, i.e. ".../df/hack/".

  • dfhack.getSavePath()

    Returns the path to the current save directory, or nil if no save loaded.

  • dfhack.getTickCount()

    Returns the tick count in ms, exactly as DF ui uses.

  • dfhack.isWorldLoaded()

    Checks if the world is loaded.

  • dfhack.isMapLoaded()

    Checks if the world and map are loaded.

  • dfhack.TranslateName(name[,in_english,only_last_name])

    Convert a language_name or only the last name part to string.

  • dfhack.df2utf(string)

    Convert a string from DF’s CP437 encoding to UTF-8.

  • dfhack.df2console()

    Convert a string from DF’s CP437 encoding to the correct encoding for the DFHack console.


When printing CP437-encoded text to the console (for example, names returned from dfhack.TranslateName()), use print(dfhack.df2console(text)) to ensure proper display on all platforms.

  • dfhack.utf2df(string)

    Convert a string from UTF-8 to DF’s CP437 encoding.

  • dfhack.toSearchNormalized(string)

    Replace non-ASCII alphabetic characters in a CP437-encoded string with their nearest ASCII equivalents, if possible, and returns a CP437-encoded string. Note that the returned string may be longer than the input string. For example, ä is replaced with a, and æ is replaced with ae.

  • dfhack.run_command(command[, ...])

    Run an arbitrary DFHack command, with the core suspended, and send output to the DFHack console. The command can be passed as a table, multiple string arguments, or a single string argument (not recommended - in this case, the usual DFHack console tokenization is used).

    A command_result constant starting with CR_ is returned, where CR_OK indicates success.

    The following examples are equivalent:

    dfhack.run_command({'ls', '-a'})
    dfhack.run_command('ls', '-a')
    dfhack.run_command('ls -a')  -- not recommended
  • dfhack.run_command_silent(command[, ...])

    Similar to run_command(), but instead of printing to the console, returns an output, command_result pair. output is a single string - see dfhack.internal.runCommand() to obtain colors as well.

Gui module

  • dfhack.gui.getCurViewscreen([skip_dismissed])

    Returns the topmost viewscreen. If skip_dismissed is true, ignores screens already marked to be removed.

  • dfhack.gui.getFocusString(viewscreen)

    Returns a string representation of the current focus position in the ui. The string has a “screen/foo/bar/baz…” format.

  • dfhack.gui.getCurFocus([skip_dismissed])

    Returns the focus string of the current viewscreen.

  • dfhack.gui.getViewscreenByType(type [, depth])

    Returns the topmost viewscreen out of the top depth viewscreens with the specified type (e.g. df.viewscreen_titlest), or nil if none match. If depth is not specified or is less than 1, all viewscreens are checked.

General-purpose selections
  • dfhack.gui.getSelectedWorkshopJob([silent])

    When a job is selected in q mode, returns the job, else prints error unless silent and returns nil.

  • dfhack.gui.getSelectedJob([silent])

    Returns the job selected in a workshop or unit/jobs screen.

  • dfhack.gui.getSelectedUnit([silent])

    Returns the unit selected via v, k, unit/jobs, or a full-screen item view of a cage or suchlike.

  • dfhack.gui.getSelectedItem([silent])

    Returns the item selected via v ->inventory, k, t, or a full-screen item view of a container. Note that in the last case, the highlighted contained item is returned, not the container itself.

  • dfhack.gui.getSelectedBuilding([silent])

    Returns the building selected via q, t, k or i.

  • dfhack.gui.getSelectedPlant([silent])

    Returns the plant selected via k.

  • dfhack.gui.getAnyUnit(screen)

  • dfhack.gui.getAnyItem(screen)

  • dfhack.gui.getAnyBuilding(screen)

  • dfhack.gui.getAnyPlant(screen)

    Similar to the corresponding getSelected functions, but operate on the screen given instead of the current screen and always return nil silently on failure.

Fortress mode
  • dfhack.gui.getDwarfmodeViewDims()

    Returns dimensions of the main fortress mode screen. See getPanelLayout() in the gui.dwarfmode module for a more Lua-friendly version.

  • dfhack.gui.resetDwarfmodeView([pause])

    Resets the fortress mode sidebar menus and cursors to their default state. If pause is true, also pauses the game.

  • dfhack.gui.revealInDwarfmodeMap(pos)

    Centers the view on the given position, which can be a df.coord instance or a table assignable to a df.coord (see Recursive table assignment), e.g.:

    {x = 5, y = 7, z = 11}

    Returns false if unsuccessful.

  • dfhack.gui.refreshSidebar()

    Refreshes the fortress mode sidebar. This can be useful when making changes to the map, for example, because DF only updates the sidebar when the cursor position changes.

  • dfhack.gui.inRenameBuilding()

    Returns true if a building is being renamed.

  • dfhack.gui.writeToGamelog(text)

    Writes a string to gamelog.txt without doing an announcement.

  • dfhack.gui.makeAnnouncement(type,flags,pos,text,color[,is_bright])

    Adds an announcement with given announcement_type, text, color, and brightness. The is_bright boolean actually seems to invert the brightness.

    The announcement is written to gamelog.txt. The announcement_flags argument provides a custom set of announcements.txt options, which specify if the message should actually be displayed in the announcement list, and whether to recenter or show a popup.

    Returns the index of the new announcement in df.global.world.status.reports, or -1.

  • dfhack.gui.addCombatReport(unit,slot,report_index)

    Adds the report with the given index (returned by makeAnnouncement) to the specified group of the given unit. Returns true on success.

  • dfhack.gui.addCombatReportAuto(unit,flags,report_index)

    Adds the report with the given index to the appropriate group(s) of the given unit, as requested by the flags.

  • dfhack.gui.showAnnouncement(text,color[,is_bright])

    Adds a regular announcement with given text, color, and brightness. The is_bright boolean actually seems to invert the brightness.

  • dfhack.gui.showZoomAnnouncement(type,pos,text,color[,is_bright])

    Like above, but also specifies a position you can zoom to from the announcement menu.

  • dfhack.gui.showPopupAnnouncement(text,color[,is_bright])

    Pops up a titan-style modal announcement window.

  • dfhack.gui.showAutoAnnouncement(type,pos,text,color[,is_bright,unit1,unit2])

    Uses the type to look up options from announcements.txt, and calls the above operations accordingly. The units are used to call addCombatReportAuto.

  • dfhack.gui.getDepthAt(x, y)

    Returns the distance from the z-level of the tile at map coordinates (x, y) to the closest ground z-level below. Defaults to 0, unless overriden by plugins.

Job module

  • dfhack.job.cloneJobStruct(job)

    Creates a deep copy of the given job.

  • dfhack.job.printJobDetails(job)

    Prints info about the job.

  • dfhack.job.printItemDetails(jobitem,idx)

    Prints info about the job item.

  • dfhack.job.getGeneralRef(job, type)

    Searches for a general_ref with the given type.

  • dfhack.job.getSpecificRef(job, type)

    Searches for a specific_ref with the given type.

  • dfhack.job.getHolder(job)

    Returns the building holding the job.

  • dfhack.job.getWorker(job)

    Returns the unit performing the job.

  • dfhack.job.setJobCooldown(building,worker,cooldown)

    Prevent the worker from taking jobs at the specified workshop for the specified cooldown period (in ticks). This doesn’t decrease the cooldown period in any circumstances.

  • dfhack.job.removeWorker(job,cooldown)

    Removes the worker from the specified workshop job, and sets the cooldown period (using the same logic as setJobCooldown). Returns true on success.

  • dfhack.job.checkBuildingsNow()

    Instructs the game to check buildings for jobs next frame and assign workers.

  • dfhack.job.checkDesignationsNow()

    Instructs the game to check designations for jobs next frame and assign workers.

  • dfhack.job.is_equal(job1,job2)

    Compares important fields in the job and nested item structures.

  • dfhack.job.is_item_equal(job_item1,job_item2)

    Compares important fields in the job item structures.

  • dfhack.job.linkIntoWorld(job,new_id)

    Adds job into df.global.job_list, and if new_id is true, then also sets its id and increases df.global.job_next_id

  • dfhack.job.listNewlyCreated(first_id)

    Returns the current value of df.global.job_next_id, and if there are any jobs with first_id <= id < job_next_id, a lua list containing them.

  • dfhack.job.isSuitableItem(job_item, item_type, item_subtype)

    Does basic sanity checks to verify if the suggested item type matches the flags in the job item.

  • dfhack.job.isSuitableMaterial(job_item, mat_type, mat_index, item_type)

    Likewise, if replacing material.

  • dfhack.job.getName(job)

    Returns the job’s description, as seen in the Units and Jobs screens.

Units module

  • dfhack.units.getPosition(unit)

    Returns true x,y,z of the unit, or nil if invalid; may be not equal to unit.pos if caged.

  • dfhack.getUnitsInBox(x1,y1,z1,x2,y2,z2[,filter])

    Returns a table of all units within the specified coordinates. If the filter argument is given, only units where filter(unit) returns true will be included. Note that pos2xyz() cannot currently be used to convert coordinate objects to the arguments required by this function.

  • dfhack.units.getGeneralRef(unit, type)

    Searches for a general_ref with the given type.

  • dfhack.units.getSpecificRef(unit, type)

    Searches for a specific_ref with the given type.

  • dfhack.units.getContainer(unit)

    Returns the container (cage) item or nil.

  • dfhack.units.setNickname(unit,nick)

    Sets the unit’s nickname properly.

  • dfhack.units.getVisibleName(unit)

    Returns the language_name object visible in game, accounting for false identities.

  • dfhack.units.getIdentity(unit)

    Returns the false identity of the unit if it has one, or nil.

  • dfhack.units.getNemesis(unit)

    Returns the nemesis record of the unit if it has one, or nil.

  • dfhack.units.isHidingCurse(unit)

    Checks if the unit hides improved attributes from its curse.

  • dfhack.units.getPhysicalAttrValue(unit, attr_type)

  • dfhack.units.getMentalAttrValue(unit, attr_type)

    Computes the effective attribute value, including curse effect.

  • dfhack.units.isCrazed(unit)

  • dfhack.units.isOpposedToLife(unit)

  • dfhack.units.hasExtravision(unit)

  • dfhack.units.isBloodsucker(unit)

    Simple checks of caste attributes that can be modified by curses.

  • dfhack.units.getMiscTrait(unit, type[, create])

    Finds (or creates if requested) a misc trait object with the given id.

  • dfhack.units.isActive(unit)

    The unit is active (alive and on the map).

  • dfhack.units.isAlive(unit)

    The unit isn’t dead or undead.

  • dfhack.units.isDead(unit)

    The unit is completely dead and passive, or a ghost. Equivalent to dfhack.units.isKilled(unit) or dfhack.units.isGhost(unit).

  • dfhack.units.isKilled(unit)

    The unit has been killed.

  • dfhack.units.isGhost(unit)

    The unit is a ghost.

  • dfhack.units.isSane(unit)

    The unit is capable of rational action, i.e. not dead, insane, zombie, or active werewolf.

  • dfhack.units.isDwarf(unit)

    The unit is of the correct race of the fortress.

  • dfhack.units.isCitizen(unit)

    The unit is an alive sane citizen of the fortress; wraps the same checks the game uses to decide game-over by extinction.

  • dfhack.units.isVisible(unit)

    The unit is visible on the map.

  • dfhack.units.getAge(unit[,true_age])

    Returns the age of the unit in years as a floating-point value. If true_age is true, ignores false identities.

  • dfhack.units.isValidLabor(unit, unit_labor)

    Returns whether the indicated labor is settable for the given unit.

  • dfhack.units.setLaborValidity(unit_labor, isValid)

    Sets the given labor to the given (boolean) validity for all units that are part of your fortress civilization. Valid labors are allowed to be toggled in the in-game labor management screens (including DFHack’s labor manipulator screen).

  • dfhack.units.getNominalSkill(unit, skill[, use_rust])

    Retrieves the nominal skill level for the given unit. If use_rust is true, subtracts the rust penalty.

  • dfhack.units.getEffectiveSkill(unit, skill)

    Computes the effective rating for the given skill, taking into account exhaustion, pain etc.

  • dfhack.units.getExperience(unit, skill[, total])

    Returns the experience value for the given skill. If total is true, adds experience implied by the current rating.

  • dfhack.units.computeMovementSpeed(unit)

    Computes number of frames * 100 it takes the unit to move in its current state of mind and body.

  • dfhack.units.computeSlowdownFactor(unit)

    Meandering and floundering in liquid introduces additional slowdown. It is random, but the function computes and returns the expected mean factor as a float.

  • dfhack.units.getNoblePositions(unit)

    Returns a list of tables describing noble position assignments, or nil. Every table has fields entity, assignment and position.

  • dfhack.units.getProfessionName(unit[,ignore_noble,plural])

    Retrieves the profession name using custom profession, noble assignments or raws. The ignore_noble boolean disables the use of noble positions.

  • dfhack.units.getCasteProfessionName(race,caste,prof_id[,plural])

    Retrieves the profession name for the given race/caste using raws.

  • dfhack.units.getProfessionColor(unit[,ignore_noble])

    Retrieves the color associated with the profession, using noble assignments or raws. The ignore_noble boolean disables the use of noble positions.

  • dfhack.units.getCasteProfessionColor(race,caste,prof_id)

    Retrieves the profession color for the given race/caste using raws.

  • dfhack.units.getGoalType(unit[,goalIndex])

    Retrieves the goal type of the dream that the given unit has. By default the goal of the first dream is returned. The goalIndex parameter may be used to retrieve additional dream goals. Currently only one dream per unit is supported by Dwarf Fortress. Support for multiple dreams may be added in future versions of Dwarf Fortress.

  • dfhack.units.getGoalName(unit[,goalIndex])

    Retrieves the short name describing the goal of the dream that the given unit has. By default the goal of the first dream is returned. The goalIndex parameter may be used to retrieve additional dream goals. Currently only one dream per unit is supported by Dwarf Fortress. Support for multiple dreams may be added in future versions of Dwarf Fortress.

  • dfhack.units.isGoalAchieved(unit[,goalIndex])

    Checks if given unit has achieved the goal of the dream. By default the status of the goal of the first dream is returned. The goalIndex parameter may be used to check additional dream goals. Currently only one dream per unit is supported by Dwarf Fortress. Support for multiple dreams may be added in future versions of Dwarf Fortress.

  • dfhack.units.getStressCategory(unit)

    Returns a number from 0-6 indicating stress. 0 is most stressed; 6 is least. Note that 0 is guaranteed to remain the most stressed but 6 could change in the future.

  • dfhack.units.getStressCategoryRaw(stress_level)

    Identical to getStressCategory but takes a raw stress level instead of a unit.

  • dfhack.units.getStressCutoffs()

    Returns a table of the cutoffs used by the above stress level functions.

Items module

  • dfhack.items.getPosition(item)

    Returns true x,y,z of the item, or nil if invalid; may be not equal to item.pos if in inventory.

  • dfhack.items.getBookTitle(item)

    Returns the title of the “book” item, or an empty string if the item isn’t a “book” or it doesn’t have a title. A “book” is a codex or a tool item that has page or writings improvements, such as scrolls and quires.

  • dfhack.items.getDescription(item, type[, decorate])

    Returns the string description of the item, as produced by the getItemDescription method. If decorate is true, also adds markings for quality and improvements.

  • dfhack.items.getGeneralRef(item, type)

    Searches for a general_ref with the given type.

  • dfhack.items.getSpecificRef(item, type)

    Searches for a specific_ref with the given type.

  • dfhack.items.getOwner(item)

    Returns the owner unit or nil.

  • dfhack.items.setOwner(item,unit)

    Replaces the owner of the item. If unit is nil, removes ownership. Returns false in case of error.

  • dfhack.items.getContainer(item)

    Returns the container item or nil.

  • dfhack.items.getContainedItems(item)

    Returns a list of items contained in this one.

  • dfhack.items.getHolderBuilding(item)

    Returns the holder building or nil.

  • dfhack.items.getHolderUnit(item)

    Returns the holder unit or nil.

  • dfhack.items.moveToGround(item,pos)

    Move the item to the ground at position. Returns false if impossible.

  • dfhack.items.moveToContainer(item,container)

    Move the item to the container. Returns false if impossible.

  • dfhack.items.moveToBuilding(item,building[,use_mode[,force_in_building])

    Move the item to the building. Returns false if impossible.

    use_mode defaults to 0. If set to 2, the item will be treated as part of the building.

    If force_in_building is true, the item will be considered to be stored by the building (used for items temporarily used in traps in vanilla DF)

  • dfhack.items.moveToInventory(item,unit,use_mode,body_part)

    Move the item to the unit inventory. Returns false if impossible.

  • dfhack.items.remove(item[, no_uncat])

    Removes the item, and marks it for garbage collection unless no_uncat is true.

  • dfhack.items.makeProjectile(item)

    Turns the item into a projectile, and returns the new object, or nil if impossible.

  • dfhack.items.isCasteMaterial(item_type)

    Returns true if this item type uses a creature/caste pair as its material.

  • dfhack.items.getSubtypeCount(item_type)

    Returns the number of raw-defined subtypes of the given item type, or -1 if not applicable.

  • dfhack.items.getSubtypeDef(item_type, subtype)

    Returns the raw definition for the given item type and subtype, or nil if invalid.

  • dfhack.items.getItemBaseValue(item_type, subtype, material, mat_index)

    Calculates the base value for an item of the specified type and material.

  • dfhack.items.getValue(item)

    Calculates the Basic Value of an item, as seen in the View Item screen.

  • dfhack.items.createItem(item_type, item_subtype, mat_type, mat_index, unit)

    Creates an item, similar to the createitem plugin.

  • dfhack.items.checkMandates(item)

    Returns true if the item is free from mandates, or false if mandates prevent trading the item.

  • dfhack.items.canTrade(item)

    Checks whether the item can be traded.

  • dfhack.items.canTradeWithContents(item)

    Checks whether the item and all items it contains, if any, can be traded.

  • dfhack.items.isRouteVehicle(item)

    Checks whether the item is an assigned hauling vehicle.

  • dfhack.items.isSquadEquipment(item)

    Checks whether the item is assigned to a squad.

Maps module

  • dfhack.maps.getSize()

    Returns map size in blocks: x, y, z

  • dfhack.maps.getTileSize()

    Returns map size in tiles: x, y, z

  • dfhack.maps.getBlock(x,y,z)

    Returns a map block object for given x,y,z in local block coordinates.

  • dfhack.maps.isValidTilePos(coords), or isValidTilePos(x,y,z)

    Checks if the given df::coord or x,y,z in local tile coordinates are valid.

  • dfhack.maps.isTileVisible(coords), or isTileVisible(x,y,z)

    Checks if the given df::coord or x,y,z in local tile coordinates is visible.

  • dfhack.maps.getTileBlock(coords), or getTileBlock(x,y,z)

    Returns a map block object for given df::coord or x,y,z in local tile coordinates.

  • dfhack.maps.ensureTileBlock(coords), or ensureTileBlock(x,y,z)

    Like getTileBlock, but if the block is not allocated, try creating it.

  • dfhack.maps.getTileType(coords), or getTileType(x,y,z)

    Returns the tile type at the given coordinates, or nil if invalid.

  • dfhack.maps.getTileFlags(coords), or getTileFlags(x,y,z)

    Returns designation and occupancy references for the given coordinates, or nil, nil if invalid.

  • dfhack.maps.getRegionBiome(region_coord2d), or getRegionBiome(x,y)

    Returns the biome info struct for the given global map region.

  • dfhack.maps.enableBlockUpdates(block[,flow,temperature])

    Enables updates for liquid flow or temperature, unless already active.

  • dfhack.maps.spawnFlow(pos,type,mat_type,mat_index,dimension)

    Spawns a new flow (i.e. steam/mist/dust/etc) at the given pos, and with the given parameters. Returns it, or nil if unsuccessful.

  • dfhack.maps.getGlobalInitFeature(index)

    Returns the global feature object with the given index.

  • dfhack.maps.getLocalInitFeature(region_coord2d,index)

    Returns the local feature object with the given region coords and index.

  • dfhack.maps.getTileBiomeRgn(coords), or getTileBiomeRgn(x,y,z)

    Returns x, y for use with getRegionBiome.

  • dfhack.maps.canWalkBetween(pos1, pos2)

    Checks if a dwarf may be able to walk between the two tiles, using a pathfinding cache maintained by the game.


    This cache is only updated when the game is unpaused, and thus can get out of date if doors are forbidden or unforbidden, or tools like liquids or tiletypes are used. It also cannot possibly take into account anything that depends on the actual units, like burrows, or the presence of invaders.

  • dfhack.maps.hasTileAssignment(tilemask)

    Checks if the tile_bitmask object is not nil and contains any set bits; returns true or false.

  • dfhack.maps.getTileAssignment(tilemask,x,y)

    Checks if the tile_bitmask object is not nil and has the relevant bit set; returns true or false.

  • dfhack.maps.setTileAssignment(tilemask,x,y,enable)

    Sets the relevant bit in the tile_bitmask object to the enable argument.

  • dfhack.maps.resetTileAssignment(tilemask[,enable])

    Sets all bits in the mask to the enable argument.

Burrows module

  • dfhack.burrows.findByName(name)

    Returns the burrow pointer or nil.

  • dfhack.burrows.clearUnits(burrow)

    Removes all units from the burrow.

  • dfhack.burrows.isAssignedUnit(burrow,unit)

    Checks if the unit is in the burrow.

  • dfhack.burrows.setAssignedUnit(burrow,unit,enable)

    Adds or removes the unit from the burrow.

  • dfhack.burrows.clearTiles(burrow)

    Removes all tiles from the burrow.

  • dfhack.burrows.listBlocks(burrow)

    Returns a table of map block pointers.

  • dfhack.burrows.isAssignedTile(burrow,tile_coord)

    Checks if the tile is in burrow.

  • dfhack.burrows.setAssignedTile(burrow,tile_coord,enable)

    Adds or removes the tile from the burrow. Returns false if invalid coords.

  • dfhack.burrows.isAssignedBlockTile(burrow,block,x,y)

    Checks if the tile within the block is in burrow.

  • dfhack.burrows.setAssignedBlockTile(burrow,block,x,y,enable)

    Adds or removes the tile from the burrow. Returns false if invalid coords.

Buildings module

  • dfhack.buildings.getGeneralRef(building, type)

    Searches for a general_ref with the given type.

  • dfhack.buildings.getSpecificRef(building, type)

    Searches for a specific_ref with the given type.

  • dfhack.buildings.setOwner(item,unit)

    Replaces the owner of the building. If unit is nil, removes ownership. Returns false in case of error.

  • dfhack.buildings.getSize(building)

    Returns width, height, centerx, centery.

  • dfhack.buildings.findAtTile(pos), or findAtTile(x,y,z)

    Scans the buildings for the one located at the given tile. Does not work on civzones. Warning: linear scan if the map tile indicates there are buildings at it.

  • dfhack.buildings.findCivzonesAt(pos), or findCivzonesAt(x,y,z)

    Scans civzones, and returns a lua sequence of those that touch the given tile, or nil if none.

  • dfhack.buildings.getCorrectSize(width, height, type, subtype, custom, direction)

    Computes correct dimensions for the specified building type and orientation, using width and height for flexible dimensions. Returns is_flexible, width, height, center_x, center_y.

  • dfhack.buildings.checkFreeTiles(pos,size[,extents,change_extents,allow_occupied,allow_wall])

    Checks if the rectangle defined by pos and size, and possibly extents, can be used for placing a building. If change_extents is true, bad tiles are removed from extents. If allow_occupied, the occupancy test is skipped. Set allow_wall to true if the building is unhindered by walls (such as an activity zone).

  • dfhack.buildings.countExtentTiles(extents,defval)

    Returns the number of tiles included by extents, or defval.

  • dfhack.buildings.containsTile(building, x, y[, room])

    Checks if the building contains the specified tile, either directly, or as room.

  • dfhack.buildings.hasSupport(pos,size)

    Checks if a bridge constructed at specified position would have support from terrain, and thus won’t collapse if retracted.

  • dfhack.buildings.getStockpileContents(stockpile)

    Returns a list of items stored on the given stockpile. Ignores empty bins, barrels, and wheelbarrows assigned as storage and transport for that stockpile.

  • dfhack.buildings.getCageOccupants(cage)

    Returns a list of units in the given built cage. Note that this is different from the list of units assigned to the cage, which can be accessed with cage.assigned_units.


Low-level building creation functions:

  • dfhack.buildings.allocInstance(pos, type, subtype, custom)

    Creates a new building instance of given type, subtype and custom type, at specified position. Returns the object, or nil in case of an error.

  • dfhack.buildings.setSize(building, width, height, direction)

    Configures an object returned by allocInstance, using specified parameters wherever appropriate. If the building has fixed size along any dimension, the corresponding input parameter will be ignored. Returns false if the building cannot be placed, or true, width, height, rect_area, true_area. Returned width and height are the final values used by the building; true_area is less than rect_area if any tiles were removed from designation. You can specify a non-rectangular designation for building types that support extents by setting the room.extents bitmap before calling this function. The extents will be reset, however, if the size returned by this function doesn’t match the input size parameter.

  • dfhack.buildings.constructAbstract(building)

    Links a fully configured object created by allocInstance into the world. The object must be an abstract building, i.e. a stockpile or civzone. Returns true, or false if impossible.

  • dfhack.buildings.constructWithItems(building, items)

    Links a fully configured object created by allocInstance into the world for construction, using a list of specific items as material. Returns true, or false if impossible.

  • dfhack.buildings.constructWithFilters(building, job_items)

    Links a fully configured object created by allocInstance into the world for construction, using a list of job_item filters as inputs. Returns true, or false if impossible. Filter objects are claimed and possibly destroyed in any case. Use a negative quantity field value to auto-compute the amount from the size of the building.

  • dfhack.buildings.deconstruct(building)

    Destroys the building, or queues a deconstruction job. Returns true if the building was destroyed and deallocated immediately.

  • dfhack.buildings.markedForRemoval(building)

    Returns true if the building is marked for removal (with x), false otherwise.

  • dfhack.buildings.getRoomDescription(building[, unit])

    If the building is a room, returns a description including quality modifiers, e.g. “Royal Bedroom”. Otherwise, returns an empty string.

    The unit argument is passed through to DF and may modify the room’s value depending on the unit given.


More high-level functions are implemented in lua and can be loaded by require('dfhack.buildings'). See hack/lua/dfhack/buildings.lua.

Among them are:

  • dfhack.buildings.getFiltersByType(argtable,type,subtype,custom)

    Returns a sequence of lua structures, describing input item filters suitable for the specified building type, or nil if unknown or invalid. The returned sequence is suitable for use as the job_items argument of constructWithFilters. Uses tables defined in buildings.lua.

    Argtable members material (the default name), bucket, barrel, chain, mechanism, screw, pipe, anvil, weapon are used to augment the basic attributes with more detailed information if the building has input items with the matching name (see the tables for naming details). Note that it is impossible to override any properties this way, only supply those that are not mentioned otherwise; one exception is that flags2.non_economic is automatically cleared if an explicit material is specified.

  • dfhack.buildings.constructBuilding{...}

    Creates a building in one call, using options contained in the argument table. Returns the building, or nil, error.


    Despite the name, unless the building is abstract, the function creates it in an ‘unconstructed’ stage, with a queued in-game job that will actually construct it. I.e. the function replicates programmatically what can be done through the construct building menu in the game ui, except that it does less environment constraint checking.

    The following options can be used:

    • pos = coordinates, or x = ..., y = ..., z = ...

      Mandatory. Specifies the left upper corner of the building.

    • type = df.building_type.FOO, subtype = ..., custom = ...

      Mandatory. Specifies the type of the building. Obviously, subtype and custom are only expected if the type requires them.

    • fields = { ... }

      Initializes fields of the building object after creation with df.assign. If room.extents is assigned this way and this function returns with error, the memory allocated for the extents is freed.

    • width = ..., height = ..., direction = ...

      Sets size and orientation of the building. If it is fixed-size, specified dimensions are ignored.

    • full_rectangle = true

      For buildings like stockpiles or farm plots that can normally accomodate individual tile exclusion, forces an error if any tiles within the specified width*height are obstructed.

    • items = { item, item ... }, or filters = { {...}, {...}... }

      Specifies explicit items or item filters to use in construction. It is the job of the user to ensure they are correct for the building type.

    • abstract = true

      Specifies that the building is abstract and does not require construction. Required for stockpiles and civzones; an error otherwise.

    • material = {...}, mechanism = {...}, ...

      If none of items, filter, or abstract is used, the function uses getFiltersByType to compute the input item filters, and passes the argument table through. If no filters can be determined this way, constructBuilding throws an error.

Constructions module

  • dfhack.constructions.designateNew(pos,type,item_type,mat_index)

    Designates a new construction at given position. If there already is a planned but not completed construction there, changes its type. Returns true, or false if obstructed. Note that designated constructions are technically buildings.

  • dfhack.constructions.designateRemove(pos), or designateRemove(x,y,z)

    If there is a construction or a planned construction at the specified coordinates, designates it for removal, or instantly cancels the planned one. Returns true, was_only_planned if removed; or false if none found.

Kitchen module

  • dfhack.kitchen.findExclusion(type, item_type, item_subtype, mat_type, mat_index)

    Finds a kitchen exclusion in the vectors in df.global.ui.kitchen. Returns -1 if not found.

    • type is a df.kitchen_exc_type, i.e. df.kitchen_exc_type.Cook or df.kitchen_exc_type.Brew.
    • item_type is a df.item_type
    • item_subtype, mat_type, and mat_index are all numeric
  • dfhack.kitchen.addExclusion(type, item_type, item_subtype, mat_type, mat_index)

  • dfhack.kitchen.removeExclusion(type, item_type, item_subtype, mat_type, mat_index)

    Adds or removes a kitchen exclusion, using the same parameters as findExclusion. Both return true on success and false on failure, e.g. when adding an exclusion that already exists or removing one that does not.

Screen API

The screen module implements support for drawing to the tiled screen of the game. Note that drawing only has any effect when done from callbacks, so it can only be feasibly used in the core context.

Basic painting functions

Common parameters to these functions include:

  • x, y: screen coordinates in tiles; the upper left corner of the screen is x = 0, y = 0
  • pen: a pen object
  • map: a boolean indicating whether to draw to a separate map buffer (defaults to false, which is suitable for off-map text or a screen that hides the map entirely). Note that only third-party plugins like TWBT currently implement a separate map buffer. If no such plugins are enabled, passing true has no effect. However, this parameter should still be used to ensure that scripts work properly with such plugins.


  • dfhack.screen.getWindowSize()

    Returns width, height of the screen.

  • dfhack.screen.getMousePos()

    Returns x,y of the tile the mouse is over.

  • dfhack.screen.inGraphicsMode()

    Checks if [GRAPHICS:YES] was specified in init.

  • dfhack.screen.paintTile(pen,x,y[,char,tile,map])

    Paints a tile using given parameters. See below for a description of pen.

    Returns false on error, e.g. if coordinates are out of bounds

  • dfhack.screen.readTile(x,y[,map])

    Retrieves the contents of the specified tile from the screen buffers. Returns a pen object, or nil if invalid or TrueType.

  • dfhack.screen.paintString(pen,x,y,text[,map])

    Paints the string starting at x,y. Uses the string characters in sequence to override the ch field of pen.

    Returns true if painting at least one character succeeded.

  • dfhack.screen.fillRect(pen,x1,y1,x2,y2[,map])

    Fills the rectangle specified by the coordinates with the given pen. Returns true if painting at least one character succeeded.

  • dfhack.screen.findGraphicsTile(pagename,x,y)

    Finds a tile from a graphics set (i.e. the raws used for creatures), if in graphics mode and loaded.

    Returns: tile, tile_grayscale, or nil if not found. The values can then be used for the tile field of pen structures.

  • dfhack.screen.clear()

    Fills the screen with blank background.

  • dfhack.screen.invalidate()

    Requests repaint of the screen by setting a flag. Unlike other functions in this section, this may be used at any time.

  • dfhack.screen.getKeyDisplay(key)

    Returns the string that should be used to represent the given logical keybinding on the screen in texts like “press Key to …”.

  • dfhack.screen.keyToChar(key)

    Returns the integer character code of the string input character represented by the given logical keybinding, or nil if not a string input key.

  • dfhack.screen.charToKey(charcode)

    Returns the keybinding representing the given string input character, or nil if impossible.


The pen argument used by dfhack.screen functions may be represented by a table with the following possible fields:

Provides the ordinary tile character, as either a 1-character string or a number. Can be overridden with the char function parameter.
Foreground color for the ordinary tile. Defaults to COLOR_GREY (7).
Background color for the ordinary tile. Defaults to COLOR_BLACK (0).
Bright/bold text flag. If nil, computed based on (fg & 8); fg is masked to 3 bits. Otherwise should be true/false.
Graphical tile id. Ignored unless [GRAPHICS:YES] was in init.txt.
tile_color = true
Specifies that the tile should be shaded with fg/bg.
tile_fg, tile_bg
If specified, overrides tile_color and supplies shading colors directly.

Alternatively, it may be a pre-parsed native object with the following API:

  • dfhack.pen.make(base[,pen_or_fg,bg,bold])

    Creates a new pre-parsed pen by combining its arguments according to the following rules:

    1. The base argument may be a pen object, a pen table as specified above, or a single color value. In the single value case, it is split into fg and bold properties, and others are initialized to 0. This argument will be converted to a pre-parsed object and returned if there are no other arguments.
    2. If the pen_or_fg argument is specified as a table or object, it completely replaces the base, and is returned instead of it.
    3. Otherwise, the non-nil subset of the optional arguments is used to update the fg, bg and bold properties of the base. If the bold flag is nil, but pen_or_fg is a number, bold is deduced from it like in the simple base case.

    This function always returns a new pre-parsed pen, or nil.

  • dfhack.pen.parse(base[,pen_or_fg,bg,bold])

    Exactly like the above function, but returns base or pen_or_fg directly if they are already a pre-parsed native object.

  • pen.property, pen.property = value, pairs(pen)

    Pre-parsed pens support reading and setting their properties, but don’t behave exactly like a simple table would; for instance, assigning to pen.tile_color also resets pen.tile_fg and pen.tile_bg to nil.

Screen management

In order to actually be able to paint to the screen, it is necessary to create and register a viewscreen (basically a modal dialog) with the game.


As a matter of policy, in order to avoid user confusion, all interface screens added by dfhack should bear the “DFHack” signature.

Screens are managed with the following functions:

  • dfhack.screen.show(screen[,below])

    Displays the given screen, possibly placing it below a different one. The screen must not be already shown. Returns true if success.

  • dfhack.screen.dismiss(screen[,to_first])

    Marks the screen to be removed when the game enters its event loop. If to_first is true, all screens up to the first one will be deleted.

  • dfhack.screen.isDismissed(screen)

    Checks if the screen is already marked for removal.

Apart from a native viewscreen object, these functions accept a table as a screen. In this case, show creates a new native viewscreen that delegates all processing to methods stored in that table.


  • The gui.Screen class provides stubs for all of the functions listed below, and its use is recommended
  • Lua-implemented screens are only supported in the core context.

Supported callbacks and fields are:

  • screen._native

    Initialized by show with a reference to the backing viewscreen object, and removed again when the object is deleted.

  • function screen:onShow()

    Called by dfhack.screen.show if successful.

  • function screen:onDismiss()

    Called by dfhack.screen.dismiss if successful.

  • function screen:onDestroy()

    Called from the destructor when the viewscreen is deleted.

  • function screen:onResize(w, h)

    Called before onRender or onIdle when the window size has changed.

  • function screen:onRender()

    Called when the viewscreen should paint itself. This is the only context where the above painting functions work correctly.

    If omitted, the screen is cleared; otherwise it should do that itself. In order to make a see-through dialog, call self._native.parent:render().

  • function screen:onIdle()

    Called every frame when the screen is on top of the stack.

  • function screen:onHelp()

    Called when the help keybinding is activated (usually ‘?’).

  • function screen:onInput(keys)

    Called when keyboard or mouse events are available. If any keys are pressed, the keys argument is a table mapping them to true. Note that this refers to logical keybingings computed from real keys via options; if multiple interpretations exist, the table will contain multiple keys.

    The table also may contain special keys:


    Maps to an integer in range 0-255. Duplicates a separate “STRING_A???” code for convenience.


    If the left or right mouse button is being pressed.


    If the left or right mouse button was just pressed.

    If this method is omitted, the screen is dismissed on receival of the LEAVESCREEN key.

  • function screen:onGetSelectedUnit()

  • function screen:onGetSelectedItem()

  • function screen:onGetSelectedJob()

  • function screen:onGetSelectedBuilding()

    Implement these to provide a return value for the matching dfhack.gui.getSelected... function.

PenArray class

Screens that require significant computation in their onRender() method can use a dfhack.penarray instance to cache their output.

  • dfhack.penarray.new(w, h)

    Creates a new penarray instance with an internal buffer of w * h tiles. These dimensions currently cannot be changed after a penarray is instantiated.

  • penarray:clear()

    Clears the internal buffer, similar to dfhack.screen.clear().

  • penarray:get_dims()

    Returns the x and y dimensions of the internal buffer.

  • penarray:get_tile(x, y)

    Returns a pen corresponding to the tile at (x, y) in the internal buffer. Note that indices are 0-based.

  • penarray:set_tile(x, y, pen)

    Sets the tile at (x, y) in the internal buffer to the pen given.

  • penarray:draw(x, y, w, h, bufferx, buffery)

    Draws the contents of the internal buffer, beginning at (bufferx, buffery) and spanning w columns and h rows, to the screen starting at (x, y). Any invalid screen and buffer coordinates are skipped.

    bufferx and buffery default to 0.

Filesystem module

Most of these functions return true on success and false on failure, unless otherwise noted.

  • dfhack.filesystem.exists(path)

    Returns true if path exists.

  • dfhack.filesystem.isfile(path)

    Returns true if path exists and is a file.

  • dfhack.filesystem.isdir(path)

    Returns true if path exists and is a directory.

  • dfhack.filesystem.getcwd()

    Returns the current working directory. To retrieve the DF path, use dfhack.getDFPath() instead.

  • dfhack.filesystem.chdir(path)

    Changes the current directory to path. Use with caution.

  • dfhack.filesystem.restore_cwd()

    Restores the current working directory to what it was when DF started.

  • dfhack.filesystem.get_initial_cwd()

    Returns the value of the working directory when DF was started.

  • dfhack.filesystem.mkdir(path)

    Creates a new directory. Returns false if unsuccessful, including if path already exists.

  • dfhack.filesystem.mkdir_recursive(path)

    Creates a new directory, including any intermediate directories that don’t exist yet. Returns true if the folder was created or already existed, or false if unsuccessful.

  • dfhack.filesystem.rmdir(path)

    Removes a directory. Only works if the directory is already empty.

  • dfhack.filesystem.mtime(path)

    Returns the modification time (in seconds) of the file or directory specified by path, or -1 if path does not exist. This depends on the system clock and should only be used locally.

  • dfhack.filesystem.atime(path)

  • dfhack.filesystem.ctime(path)

    Return values vary across operating systems - return the st_atime and st_ctime fields of a C++ stat struct, respectively.

  • dfhack.filesystem.listdir(path)

    Lists files/directories in a directory. Returns {} if path does not exist. Set include_prefix to false if you don’t want the path string prepended to the returned filenames.

  • dfhack.filesystem.listdir_recursive(path [, depth = 10[, include_prefix = true]])

    Lists all files/directories in a directory and its subdirectories. All directories are listed before their contents. Returns a table with subtables of the format:

    {path: 'path to file', isdir: true|false}

    Note that listdir() returns only the base name of each directory entry, while listdir_recursive() returns the initial path and all components following it for each entry.

Console API

  • dfhack.console.clear()

    Clears the console; equivalent to the cls built-in command.

  • dfhack.console.flush()

    Flushes all output to the console. This can be useful when printing text that does not end in a newline but should still be displayed.

Internal API

These functions are intended for the use by dfhack developers, and are only documented here for completeness:

  • dfhack.internal.getPE()

    Returns the PE timestamp of the DF executable (only on Windows)

  • dfhack.internal.getMD5()

    Returns the MD5 of the DF executable (only on OS X and Linux)

  • dfhack.internal.getAddress(name)

    Returns the global address name, or nil.

  • dfhack.internal.setAddress(name, value)

    Sets the global address name. Returns the value of getAddress before the change.

  • dfhack.internal.getVTable(name)

    Returns the pre-extracted vtable address name, or nil.

  • dfhack.internal.getImageBase()

    Returns the mmap base of the executable.

  • dfhack.internal.getRebaseDelta()

    Returns the ASLR rebase offset of the DF executable.

  • dfhack.internal.adjustOffset(offset[,to_file])

    Returns the re-aligned offset, or nil if invalid. If to_file is true, the offset is adjusted from memory to file. This function returns the original value everywhere except windows.

  • dfhack.internal.getMemRanges()

    Returns a sequence of tables describing virtual memory ranges of the process.

  • dfhack.internal.patchMemory(dest,src,count)

    Like memmove below, but works even if dest is read-only memory, e.g. code. If destination overlaps a completely invalid memory region, or another error occurs, returns false.

  • dfhack.internal.patchBytes(write_table[, verify_table])

    The first argument must be a lua table, which is interpreted as a mapping from memory addresses to byte values that should be stored there. The second argument may be a similar table of values that need to be checked before writing anything.

    The function takes care to either apply all of write_table, or none of it. An empty write_table with a nonempty verify_table can be used to reasonably safely check if the memory contains certain values.

    Returns true if successful, or nil, error_msg, address if not.

  • dfhack.internal.memmove(dest,src,count)

    Wraps the standard memmove function. Accepts both numbers and refs as pointers.

  • dfhack.internal.memcmp(ptr1,ptr2,count)

    Wraps the standard memcmp function.

  • dfhack.internal.memscan(haystack,count,step,needle,nsize)

    Searches for needle of nsize bytes in haystack, using count steps of step bytes. Returns: step_idx, sum_idx, found_ptr, or nil if not found.

  • dfhack.internal.diffscan(old_data, new_data, start_idx, end_idx, eltsize[, oldval, newval, delta])

    Searches for differences between buffers at ptr1 and ptr2, as integers of size eltsize. The oldval, newval or delta arguments may be used to specify additional constraints. Returns: found_index, or nil if end reached.

  • dfhack.internal.getDir(path)

    Lists files/directories in a directory. Returns: file_names or empty table if not found. Identical to dfhack.filesystem.listdir(path).

  • dfhack.internal.strerror(errno)

    Wraps strerror() - returns a string describing a platform-specific error code

  • dfhack.internal.addScriptPath(path, search_before)

    Registers path as a script path. If search_before is passed and true, the path will be searched before the default paths (e.g. raw/scripts, hack/scripts); otherwise, it will be searched after.

    Returns true if successful or false otherwise (e.g. if the path does not exist or has already been registered).

  • dfhack.internal.removeScriptPath(path)

    Removes path from the list of script paths and returns true if successful.

  • dfhack.internal.getScriptPaths()

    Returns the list of script paths in the order they are searched, including defaults. (This can change if a world is loaded.)

  • dfhack.internal.findScript(name)

    Searches script paths for the script name and returns the path of the first file found, or nil on failure.


    This requires an extension to be specified (.lua or .rb) - use dfhack.findScript() to include the .lua extension automatically.

  • dfhack.internal.runCommand(command[, use_console])

    Runs a DFHack command with the core suspended. Used internally by the dfhack.run_command() family of functions.

    • command: either a table of strings or a single string which is parsed by the default console tokenization strategy (not recommended)
    • use_console: if true, output is sent directly to the DFHack console

    Returns a table with a status key set to a command_result constant (status = CR_OK indicates success). Additionally, if use_console is not true, enumerated table entries of the form {color, text} are included, e.g. result[1][0] is the color of the first piece of text printed (a COLOR_ constant). These entries can be iterated over with ipairs().

  • dfhack.internal.md5(string)

    Returns the MD5 hash of the given string.

  • dfhack.internal.md5File(filename[,first_kb])

    Computes the MD5 hash of the given file. Returns hash, length on success (where length is the number of bytes read from the file), or nil, error on failure.

    If the parameter first_kb is specified and evaluates to true, and the hash was computed successfully, a table containing the first 1024 bytes of the file is returned as the third return value.

  • dfhack.internal.threadid()

    Returns a numeric identifier of the current thread.

Core interpreter context

While plugins can create any number of interpreter instances, there is one special context managed by the DFHack core. It is the only context that can receive events from DF and plugins.

Core context specific functions:

  • dfhack.is_core_context

    Boolean value; true in the core context.

  • dfhack.timeout(time,mode,callback)

    Arranges for the callback to be called once the specified period of time passes. The mode argument specifies the unit of time used, and may be one of 'frames' (raw FPS), 'ticks' (unpaused FPS), 'days', 'months', 'years' (in-game time). All timers other than 'frames' are cancelled when the world is unloaded, and cannot be queued until it is loaded again. Returns the timer id, or nil if unsuccessful due to world being unloaded.

  • dfhack.timeout_active(id[,new_callback])

    Returns the active callback with the given id, or nil if inactive or nil id. If called with 2 arguments, replaces the current callback with the given value, if still active. Using timeout_active(id,nil) cancels the timer.

  • dfhack.onStateChange.foo = function(code)

    Creates a handler for state change events. Receives the same SC_ codes as plugin_onstatechange() in C++.

Event type

An event is a native object transparently wrapping a lua table, and implementing a __call metamethod. When it is invoked, it loops through the table with next and calls all contained values. This is intended as an extensible way to add listeners.

This type itself is available in any context, but only the core context has the actual events defined by C++ code.


  • dfhack.event.new()

    Creates a new instance of an event.

  • event[key] = function

    Sets the function as one of the listeners. Assign nil to remove it.


    The df.NULL key is reserved for the use by the C++ owner of the event; it is an error to try setting it.

  • #event

    Returns the number of non-nil listeners.

  • pairs(event)

    Iterates over all listeners in the table.

  • event(args...)

    Invokes all listeners contained in the event in an arbitrary order using dfhack.safecall.

Lua Modules

DFHack sets up the lua interpreter so that the built-in require function can be used to load shared lua code from hack/lua/. The dfhack namespace reference itself may be obtained via require('dfhack'), although it is initially created as a global by C++ bootstrap code.

The following module management functions are provided:

  • mkmodule(name)

    Creates an environment table for the module. Intended to be used as:

    local _ENV = mkmodule('foo')
    return _ENV

    If called the second time, returns the same table; thus providing reload support.

  • reload(name)

    Reloads a previously require-d module “name” from the file. Intended as a help for module development.

  • dfhack.BASE_G

    This variable contains the root global environment table, which is used as a base for all module and script environments. Its contents should be kept limited to the standard Lua library and API described in this document.

Global environment

A number of variables and functions are provided in the base global environment by the mandatory init file dfhack.lua:

  • Color constants

    These are applicable both for dfhack.color() and color fields in DF functions or structures:

  • State change event codes, used by dfhack.onStateChange

    Available only in the core context, as is the event itself:


  • Command result constants (equivalent to command_result in C++), used by dfhack.run_command() and related functions:


  • Functions already described above

    safecall, qerror, mkmodule, reload

  • Miscellaneous constants


    evaluate to the relevant character strings.


    is an unspecified unique token used by the class module below.

  • printall(obj)

    If the argument is a lua table or DF object reference, prints all fields.

  • printall_recurse(obj)

    If the argument is a lua table or DF object reference, prints all fields recursively.

  • copyall(obj)

    Returns a shallow copy of the table or reference as a lua table.

  • pos2xyz(obj)

    The object must have fields x, y and z. Returns them as 3 values. If obj is nil, or x is -30000 (the usual marker for undefined coordinates), returns nil.

  • xyz2pos(x,y,z)

    Returns a table with x, y and z as fields.

  • same_xyz(a,b)

    Checks if a and b have the same x, y and z fields.

  • get_path_xyz(path,i)

    Returns path.x[i], path.y[i], path.z[i].

  • pos2xy(obj), xy2pos(x,y), same_xy(a,b), get_path_xy(a,b)

    Same as above, but for 2D coordinates.

  • safe_index(obj,index...)

    Walks a sequence of dereferences, which may be represented by numbers or strings. Returns nil if any of obj or indices is nil, or a numeric index is out of array bounds.


  • utils.compare(a,b)

    Comparator function; returns -1 if a<b, 1 if a>b, 0 otherwise.

  • utils.compare_name(a,b)

    Comparator for names; compares empty string last.

  • utils.is_container(obj)

    Checks if obj is a container ref.

  • utils.make_index_sequence(start,end)

    Returns a lua sequence of numbers in start..end.

  • utils.make_sort_order(data, ordering)

    Computes a sorted permutation of objects in data, as a table of integer indices into the data sequence. Uses data.n as input length if present.

    The ordering argument is a sequence of ordering specs, represented as lua tables with following possible fields:

    ord.key = function(value)

    Computes comparison key from input data value. Not called on nil. If omitted, the comparison key is the value itself.

    ord.key_table = function(data)

    Computes a key table from the data table in one go.

    ord.compare = function(a,b)

    Comparison function. Defaults to utils.compare above. Called on non-nil keys; nil sorts last.

    ord.nil_first = true/false

    If true, nil keys are sorted first instead of last.

    ord.reverse = true/false

    If true, sort non-nil keys in descending order.

    For every comparison during sorting the specs are applied in order until an unambiguous decision is reached. Sorting is stable.

    Example of sorting a sequence by field foo:

    local spec = { key = function(v) return v.foo end }
    local order = utils.make_sort_order(data, { spec })
    local output = {}
    for i = 1,#order do output[i] = data[order[i]] end

    Separating the actual reordering of the sequence in this way enables applying the same permutation to multiple arrays. This function is used by the sort plugin.

  • for link,item in utils.listpairs(list)

    Iterates a df-list structure, for example df.global.world.job_list.

  • utils.assign(tgt, src)

    Does a recursive assignment of src into tgt. Uses df.assign if tgt is a native object ref; otherwise recurses into lua tables.

  • utils.clone(obj, deep)

    Performs a shallow, or semi-deep copy of the object as a lua table tree. The deep mode recurses into lua tables and subobjects, except pointers to other heap objects. Null pointers are represented as df.NULL. Zero-based native containers are converted to 1-based lua sequences.

  • utils.clone_with_default(obj, default, force)

    Copies the object, using the default lua table tree as a guide to which values should be skipped as uninteresting. The force argument makes it always return a non-nil value.

  • utils.parse_bitfield_int(value, type_ref)

    Given an int value, and a bitfield type in the df tree, it returns a lua table mapping the enabled bit keys to true, unless value is 0, in which case it returns nil.

  • utils.list_bitfield_flags(bitfield[, list])

    Adds all enabled bitfield keys to list or a newly-allocated empty sequence, and returns it. The bitfield argument may be nil.

  • utils.sort_vector(vector,field,cmpfun)

    Sorts a native vector or lua sequence using the comparator function. If field is not nil, applies the comparator to the field instead of the whole object.

  • utils.linear_index(vector,key[,field])

    Searches for key in the vector, and returns index, found_value, or nil if none found.

  • utils.binsearch(vector,key,field,cmpfun,min,max)

    Does a binary search in a native vector or lua sequence for key, using cmpfun and field like sort_vector. If min and max are specified, they are used as the search subrange bounds.

    If found, returns item, true, idx. Otherwise returns nil, false, insert_idx, where insert_idx is the correct insertion point.

  • utils.insert_sorted(vector,item,field,cmpfun)

    Does a binary search, and inserts item if not found. Returns did_insert, vector[idx], idx.

  • utils.insert_or_update(vector,item,field,cmpfun)

    Like insert_sorted, but also assigns the item into the vector cell if insertion didn’t happen.

    As an example, you can use this to set skill values:

    utils.insert_or_update(soul.skills, {new=true, id=..., rating=...}, 'id')

    (For an explanation of new=true, see Recursive table assignment)

  • utils.erase_sorted_key(vector,key,field,cmpfun)

    Removes the item with the given key from the list. Returns: did_erase, vector[idx], idx.

  • utils.erase_sorted(vector,item,field,cmpfun)

    Exactly like erase_sorted_key, but if field is specified, takes the key from item[field].

  • utils.call_with_string(obj,methodname,...)

    Allocates a temporary string object, calls obj:method(tmp,...), and returns the value written into the temporary after deleting it.

  • utils.getBuildingName(building)

    Returns the string description of the given building.

  • utils.getBuildingCenter(building)

    Returns an x/y/z table pointing at the building center.

  • utils.split_string(string, delimiter)

    Splits the string by the given delimiter, and returns a sequence of results.

  • utils.prompt_yes_no(prompt, default)

    Presents a yes/no prompt to the user. If default is not nil, allows just pressing Enter to submit the default choice. If the user enters 'abort', throws an error.

  • utils.prompt_input(prompt, checkfun, quit_str)

    Presents a prompt to input data, until a valid string is entered. Once checkfun(input) returns true, …, passes the values through. If the user enters the quit_str (defaults to '~~~'), throws an error.

  • utils.check_number(text)

    A prompt_input checkfun that verifies a number input.


A third-party lua table dumper module from http://lua-users.org/wiki/DataDumper. Defines one function:

  • dumper.DataDumper(value, varname, fastmode, ident, indent_step)

    Returns value converted to a string. The indent_step argument specifies the indentation step size in spaces. For the other arguments see the original documentation link above.


A third-party lua profiler module from http://lua-users.org/wiki/PepperfishProfiler. Module defines one function to create profiler objects which can be used to profile and generate report.

  • profiler.newProfiler([variant[, sampling_frequency]])

    Returns an profile object with variant either 'time' or 'call'. 'time' variant takes optional sampling_frequency parameter to select lua instruction counts between samples. Default is 'time' variant with 10*1000 frequency.

    'call' variant has much higher runtime cost which will increase the runtime of profiled code by factor of ten. For the extreme costs it provides accurate function call counts that can help locate code which takes much time in native calls.

  • obj:start()

    Resets collected statistics. Then it starts collecting new statistics.

  • obj:stop()

    Stops profile collection.

  • obj:report(outfile[, sort_by_total_time])

    Write a report from previous statistics collection to outfile. outfile should be writeable io file object (io.open or io.stdout). Passing true as second parameter sort_by_total_time switches sorting order to use total time instead of default self time order.

  • obj:prevent(function)

    Adds an ignore filter for a function. It will ignore the pointed function and all of it children.


local prof = profiler.newProfiler()



local out = io.open( "lua-profile.txt", "w+")


Implements a trivial single-inheritance class system.

  • Foo = defclass(Foo[, ParentClass])

    Defines or updates class Foo. The Foo = defclass(Foo) syntax is needed so that when the module or script is reloaded, the class identity will be preserved through the preservation of global variable values.

    The defclass function is defined as a stub in the global namespace, and using it will auto-load the class module.

  • Class.super

    This class field is set by defclass to the parent class, and allows a readable Class.super.method(self, ...) syntax for calling superclass methods.

  • Class.ATTRS { foo = xxx, bar = yyy }

    Declares certain instance fields to be attributes, i.e. auto-initialized from fields in the table used as the constructor argument. If omitted, they are initialized with the default values specified in this declaration.

    If the default value should be nil, use ATTRS { foo = DEFAULT_NIL }.

    Declaring an attribute is mostly the same as defining your init method like this:

    function Class.init(args)
        self.attr1 = args.attr1 or default1
        self.attr2 = args.attr2 or default2

    The main difference is that attributes are processed as a separate initialization step, before any init methods are called. They also make the directy relation between instance fields and constructor arguments more explicit.

  • new_obj = Class{ foo = arg, bar = arg, ... }

    Calling the class as a function creates and initializes a new instance. Initialization happens in this order:

    1. An empty instance table is created, and its metatable set.
    2. The preinit methods are called via invoke_before (see below) with the table used as argument to the class. These methods are intended for validating and tweaking that argument table.
    3. Declared ATTRS are initialized from the argument table or their default values.
    4. The init methods are called via invoke_after with the argument table. This is the main constructor method.
    5. The postinit methods are called via invoke_after with the argument table. Place code that should be called after the object is fully constructed here.

Predefined instance methods:

  • instance:assign{ foo = xxx }

    Assigns all values in the input table to the matching instance fields.

  • instance:callback(method_name, [args...])

    Returns a closure that invokes the specified method of the class, properly passing in self, and optionally a number of initial arguments too. The arguments given to the closure are appended to these.

  • instance:cb_getfield(field_name)

    Returns a closure that returns the specified field of the object when called.

  • instance:cb_setfield(field_name)

    Returns a closure that sets the specified field to its argument when called.

  • instance:invoke_before(method_name, args...)

    Navigates the inheritance chain of the instance starting from the most specific class, and invokes the specified method with the arguments if it is defined in that specific class. Equivalent to the following definition in every class:

    function Class:invoke_before(method, ...)
      if rawget(Class, method) then
        rawget(Class, method)(self, ...)
      Class.super.invoke_before(method, ...)
  • instance:invoke_after(method_name, args...)

    Like invoke_before, only the method is called after the recursive call to super, i.e. invocations happen in the parent to child order.

    These two methods are inspired by the Common Lisp before and after methods, and are intended for implementing similar protocols for certain things. The class library itself uses them for constructors.

To avoid confusion, these methods cannot be redefined.

In-game UI Library

A number of lua modules with names starting with gui are dedicated to wrapping the natives of the dfhack.screen module in a way that is easy to use. This allows relatively easily and naturally creating dialogs that integrate in the main game UI window.

These modules make extensive use of the class module, and define things ranging from the basic Painter, View and Screen classes, to fully functional predefined dialogs.


This module defines the most important classes and functions for implementing interfaces. This documents those of them that are considered stable.



    Contains the value of dfhack.screen.inGraphicsMode(), which cannot be changed without restarting the game and thus is constant during the session.


    The black pen used to clear the screen.

  • simulateInput(screen, keys...)

    This function wraps an undocumented native function that passes a set of keycodes to a screen, and is the official way to do that.

    Every argument after the initial screen may be nil, a numeric keycode, a string keycode, a sequence of numeric or string keycodes, or a mapping of keycodes to true or false. For instance, it is possible to use the table passed as argument to onInput.

  • mkdims_xy(x1,y1,x2,y2)

    Returns a table containing the arguments as fields, and also width and height that contains the rectangle dimensions.

  • mkdims_wh(x1,y1,width,height)

    Returns the same kind of table as mkdims_xy, only this time it computes x2 and y2.

  • is_in_rect(rect,x,y)

    Checks if the given point is within a rectangle, represented by a table produced by one of the mkdims functions.

  • blink_visible(delay)

    Returns true or false, with the value switching to the opposite every delay msec. This is intended for rendering blinking interface objects.

  • getKeyDisplay(keycode)

    Wraps dfhack.screen.getKeyDisplay in order to allow using strings for the keycode argument.

ViewRect class

This class represents an on-screen rectangle with an associated independent clip area rectangle. It is the base of the Painter class, and is used by Views to track their client area.

  • ViewRect{ rect = ..., clip_rect = ..., view_rect = ..., clip_view = ... }

    The constructor has the following arguments:

    rect:The mkdims rectangle in screen coordinates of the logical viewport. Defaults to the whole screen.
    clip_rect:The clip rectangle in screen coordinates. Defaults to rect.
    view_rect:A ViewRect object to copy from; overrides both rect and clip_rect.
    clip_view:A ViewRect object to intersect the specified clip area with.
  • rect:isDefunct()

    Returns true if the clip area is empty, i.e. no painting is possible.

  • rect:inClipGlobalXY(x,y)

    Checks if these global coordinates are within the clip rectangle.

  • rect:inClipLocalXY(x,y)

    Checks if these coordinates (specified relative to x1,y1) are within the clip rectangle.

  • rect:localXY(x,y)

    Converts a pair of global coordinates to local; returns x_local,y_local.

  • rect:globalXY(x,y)

    Converts a pair of local coordinates to global; returns x_global,y_global.

  • rect:viewport(x,y,w,h) or rect:viewport(subrect)

    Returns a ViewRect representing a sub-rectangle of the current one. The arguments are specified in local coordinates; the subrect argument must be a mkdims table. The returned object consists of the exact specified rectangle, and a clip area produced by intersecting it with the clip area of the original object.

Painter class

The painting natives in dfhack.screen apply to the whole screen, are completely stateless and don’t implement clipping.

The Painter class inherits from ViewRect to provide clipping and local coordinates, and tracks current cursor position and current pen. It also supports drawing to a separate map buffer if applicable (see map() below for details).

  • Painter{ ..., pen = ..., key_pen = ... }

    In addition to ViewRect arguments, Painter accepts a suggestion of the initial value for the main pen, and the keybinding pen. They default to COLOR_GREY and COLOR_LIGHTGREEN otherwise.

    There are also some convenience functions that wrap this constructor:

    • Painter.new(rect,pen)
    • Painter.new_view(view_rect,pen)
    • Painter.new_xy(x1,y1,x2,y2,pen)
    • Painter.new_wh(x1,y1,width,height,pen)
  • painter:isValidPos()

    Checks if the current cursor position is within the clip area.

  • painter:viewport(x,y,w,h)

    Like the superclass method, but returns a Painter object.

  • painter:cursor()

    Returns the current cursor x,y in screen coordinates.

  • painter:cursorX()

    Returns just the current x cursor coordinate

  • painter:cursorY()

    Returns just the current y cursor coordinate

  • painter:seek(x,y)

    Sets the current cursor position, and returns self. Either of the arguments may be nil to keep the current value.

  • painter:advance(dx,dy)

    Adds the given offsets to the cursor position, and returns self. Either of the arguments may be nil to keep the current value.

  • painter:newline([dx])

    Advances the cursor to the start of the next line plus the given x offset, and returns self.

  • painter:pen(...)

    Sets the current pen to dfhack.pen.parse(old_pen,...), and returns self.

  • painter:color(fg[,bold[,bg]])

    Sets the specified colors of the current pen and returns self.

  • painter:key_pen(...)

    Sets the current keybinding pen to dfhack.pen.parse(old_pen,...), and returns self.

  • painter:map(to_map)

    Enables or disables drawing to a separate map buffer. to_map is a boolean that will be passed as the map parameter to any dfhack.screen functions that accept it. Note that only third-party plugins like TWBT currently implement a separate map buffer; if none are enabled, this function has no effect (but should still be used to ensure proper support for such plugins). Returns self.

  • painter:clear()

    Fills the whole clip rectangle with CLEAR_PEN, and returns self.

  • painter:fill(x1,y1,x2,y2[,...]) or painter:fill(rect[,...])

    Fills the specified local coordinate rectangle with dfhack.pen.parse(cur_pen,...), and returns self.

  • painter:char([char[, ...]])

    Paints one character using char and dfhack.pen.parse(cur_pen,...); returns self. The char argument, if not nil, is used to override the ch property of the pen.

  • painter:tile([char, tile[, ...]])

    Like char() above, but also allows overriding the tile property on ad-hoc basis.

  • painter:string(text[, ...])

    Paints the string with dfhack.pen.parse(cur_pen,...); returns self.

  • painter:key(keycode[, ...])

    Paints the description of the keycode using dfhack.pen.parse(cur_key_pen,...); returns self.

  • painter:key_string(keycode, text, ...)

    A convenience wrapper around both key() and string() that prints both the specified keycode description and text, separated by :. Any extra arguments are passed directly to string(). Returns self.

Unless specified otherwise above, all Painter methods return self, in order to allow chaining them like this:


View class

This class is the common abstract base of both the stand-alone screens and common widgets to be used inside them. It defines the basic layout, rendering and event handling framework.

The class defines the following attributes:

visible:Specifies that the view should be painted.
active:Specifies that the view should receive events, if also visible.
view_id:Specifies an identifier to easily identify the view among subviews. This is reserved for implementation of top-level views, and should not be used by widgets for their internal subviews.

It also always has the following fields:

subviews:Contains a table of all subviews. The sequence part of the table is used for iteration. In addition, subviews are also indexed under their view_id, if any; see addviews() below.

These fields are computed by the layout process:

 The ViewRect represeting the client area of the parent view.
frame_rect:The mkdims rect of the outer frame in parent-local coordinates.
frame_body:The ViewRect representing the body part of the View’s own frame.

The class has the following methods:

  • view:addviews(list)

    Adds the views in the list to the subviews sequence. If any of the views in the list have view_id attributes that don’t conflict with existing keys in subviews, also stores them under the string keys. Finally, copies any non-conflicting string keys from the subviews tables of the listed views.

    Thus, doing something like this:

            view_id = 'panel',
            subviews = {
                Label{ view_id = 'label' }

    Would make the label accessible as both self.subviews.label and self.subviews.panel.subviews.label.

  • view:getWindowSize()

    Returns the dimensions of the frame_body rectangle.

  • view:getMousePos()

    Returns the mouse x,y in coordinates local to the frame_body rectangle if it is within its clip area, or nothing otherwise.

  • view:updateLayout([parent_rect])

    Recomputes layout of the view and its subviews. If no argument is given, re-uses the previous parent rect. The process goes as follows:

    1. Calls preUpdateLayout(parent_rect) via invoke_before.
    2. Uses computeFrame(parent_rect) to compute the desired frame.
    3. Calls postComputeFrame(frame_body) via invoke_after.
    4. Calls updateSubviewLayout(frame_body) to update children.
    5. Calls postUpdateLayout(frame_body) via invoke_after.
  • view:computeFrame(parent_rect) (for overriding)

    Called by updateLayout in order to compute the frame rectangle(s). Should return the mkdims rectangle for the outer frame, and optionally also for the body frame. If only one rectangle is returned, it is used for both frames, and the margin becomes zero.

  • view:updateSubviewLayout(frame_body)

    Calls updateLayout on all children.

  • view:render(painter)

    Given the parent’s painter, renders the view via the following process:

    1. Calls onRenderFrame(painter, frame_rect) to paint the outer frame.
    2. Creates a new painter using the frame_body rect.
    3. Calls onRenderBody(new_painter) to paint the client area.
    4. Calls renderSubviews(new_painter) to paint visible children.
  • view:renderSubviews(painter)

    Calls render on all visible subviews in the order they appear in the subviews sequence.

  • view:onRenderFrame(painter, rect) (for overriding)

    Called by render to paint the outer frame; by default does nothing.

  • view:onRenderBody(painter) (for overriding)

    Called by render to paint the client area; by default does nothing.

  • view:onInput(keys) (for overriding)

    Override this to handle events. By default directly calls inputToSubviews. Return a true value from this method to signal that the event has been handled and should not be passed on to more views.

  • view:inputToSubviews(keys)

    Calls onInput on all visible active subviews, iterating the subviews sequence in reverse order, so that topmost subviews get events first. Returns true if any of the subviews handled the event.

Screen class

This is a View subclass intended for use as a stand-alone dialog or screen. It adds the following methods:

  • screen:isShown()

    Returns true if the screen is currently in the game engine’s display stack.

  • screen:isDismissed()

    Returns true if the screen is dismissed.

  • screen:isActive()

    Returns true if the screen is shown and not dismissed.

  • screen:invalidate()

    Requests a repaint. Note that currently using it is not necessary, because repaints are constantly requested automatically, due to issues with native screens happening otherwise.

  • screen:renderParent()

    Asks the parent native screen to render itself, or clears the screen if impossible.

  • screen:sendInputToParent(...)

    Uses simulateInput to send keypresses to the native parent screen.

  • screen:show([parent])

    Adds the screen to the display stack with the given screen as the parent; if parent is not specified, places this one one topmost. Before calling dfhack.screen.show, calls self:onAboutToShow(parent).

  • screen:onAboutToShow(parent) (for overriding)

    Called when dfhack.screen.show is about to be called.

  • screen:onShow()

    Called by dfhack.screen.show once the screen is successfully shown.

  • screen:dismiss()

    Dismisses the screen. A dismissed screen does not receive any more events or paint requests, but may remain in the display stack for a short time until the game removes it.

  • screen:onDismiss() (for overriding)

    Called by dfhack.screen.dismiss().

  • screen:onDestroy() (for overriding)

    Called by the native code when the screen is fully destroyed and removed from the display stack. Place code that absolutely must be called whenever the screen is removed by any means here.

  • screen:onResize, screen:onRender

    Defined as callbacks for native code.

FramedScreen class

A Screen subclass that paints a visible frame around its body. Most dialogs should inherit from this class.

A framed screen has the following attributes:

frame_style:A table that defines a set of pens to draw various parts of the frame.
frame_title:A string to display in the middle of the top of the frame.
frame_width:Desired width of the client area. If nil, the screen will occupy the whole width.
frame_height:Likewise, for height.
frame_inset:The gap between the frame and the client area. Defaults to 0.
 The pen to fill in the frame with. Defaults to CLEAR_PEN.

There are the following predefined frame style tables:


    A plain grey-colored frame.


    The same frame as used by the usual full-screen DF views, like dwarfmode.


    A frame consisting of grey lines, similar to the one used by titan announcements.


This module implements some basic widgets based on the View infrastructure.

Widget class

Base of all the widgets. Inherits from View and has the following attributes:

  • frame = {...}

    Specifies the constraints on the outer frame of the widget. If omitted, the widget will occupy the whole parent rectangle.

    The frame is specified as a table with the following possible fields:

    l:gap between the left edges of the frame and the parent.
    t:gap between the top edges of the frame and the parent.
    r:gap between the right edges of the frame and the parent.
    b:gap between the bottom edges of the frame and the parent.
    w:maximum width of the frame.
    h:maximum heigth of the frame.
    xalign:X alignment of the frame.
    yalign:Y alignment of the frame.

    First the l,t,r,b fields restrict the available area for placing the frame. If w and h are not specified or larger then the computed area, it becomes the frame. Otherwise the smaller frame is placed within the are based on the xalign/yalign fields. If the align hints are omitted, they are assumed to be 0, 1, or 0.5 based on which of the l/r/t/b fields are set.

  • frame_inset = {...}

    Specifies the gap between the outer frame, and the client area. The attribute may be a simple integer value to specify a uniform inset, or a table with the following fields:

    l:left margin.
    t:top margin.
    r:right margin.
    b:bottom margin.
    x:left/right margin, if l and/or r are omitted.
    y:top/bottom margin, if t and/or b are omitted.
  • frame_background = pen

    The pen to fill the outer frame with. Defaults to no fill.

Panel class

Inherits from Widget, and intended for grouping a number of subviews.

Has attributes:

  • subviews = {}

    Used to initialize the subview list in the constructor.

  • on_render = function(painter)

    Called from onRenderBody.

Pages class

Subclass of Panel; keeps exactly one child visible.

  • Pages{ ..., selected = ... }

    Specifies which child to select initially; defaults to the first one.

  • pages:getSelected()

    Returns the selected index, child.

  • pages:setSelected(index)

    Selects the specified child, hiding the previous selected one. It is permitted to use the subview object, or its view_id as index.

EditField class

Subclass of Widget; implements a simple edit field.


text:The current contents of the field.
text_pen:The pen to draw the text with.
on_char:Input validation callback; used as on_char(new_char,text). If it returns false, the character is ignored.
on_change:Change notification callback; used as on_change(new_text,old_text).
on_submit:Enter key callback; if set the field will handle the key and call on_submit(text).
key:If specified, the field is disabled until this key is pressed. Must be given as a string.

Label class

This Widget subclass implements flowing semi-static text.

It has the following attributes:

text_pen:Specifies the pen for active text.
text_dpen:Specifies the pen for disabled text.
text_hpen:Specifies the pen for text hovered over by the mouse, if a click handler is registered.
disabled:Boolean or a callback; if true, the label is disabled.
enabled:Boolean or a callback; if false, the label is disabled.
auto_height:Sets self.frame.h from the text height.
auto_width:Sets self.frame.w from the text width.
on_click:A callback called when the label is clicked (optional)
on_rclick:A callback called when the label is right-clicked (optional)

The text itself is represented as a complex structure, and passed to the object via the text argument of the constructor, or via the setText method, as one of:

  • A simple string, possibly containing newlines.
  • A sequence of tokens.

Every token in the sequence in turn may be either a string, possibly containing newlines, or a table with the following possible fields:

  • token.text = ...

    Specifies the main text content of a token, and may be a string, or a callback returning a string.

  • token.gap = ...

    Specifies the number of character positions to advance on the line before rendering the token.

  • token.tile = pen

    Specifies a pen to paint as one tile before the main part of the token.

  • token.width = ...

    If specified either as a value or a callback, the text field is padded or truncated to the specified number.

  • token.pad_char = '?'

    If specified together with width, the padding area is filled with this character instead of just being skipped over.

  • token.key = '...'

    Specifies the keycode associated with the token. The string description of the key binding is added to the text content of the token.

  • token.key_sep = '...'

    Specifies the separator to place between the keybinding label produced by token.key, and the main text of the token. If the separator is ‘()’, the token is formatted as text..' ('..binding..')'. Otherwise it is simply binding..sep..text.

  • token.enabled, token.disabled

    Same as the attributes of the label itself, but applies only to the token.

  • token.pen, token.dpen

    Specify the pen and disabled pen to be used for the token’s text. The field may be either the pen itself, or a callback that returns it.

  • token.on_activate

    If this field is not nil, and token.key is set, the token will actually respond to that key binding unless disabled, and call this callback. Eventually this may be extended with mouse click support.

  • token.id

    Specifies a unique identifier for the token.

  • token.line, token.x1, token.x2

    Reserved for internal use.

The Label widget implements the following methods:

  • label:setText(new_text)

    Replaces the text currently contained in the widget.

  • label:itemById(id)

    Finds a token by its id field.

  • label:getTextHeight()

    Computes the height of the text.

  • label:getTextWidth()

    Computes the width of the text.

List class

The List widget implements a simple list with paging.

It has the following attributes:

text_pen:Specifies the pen for deselected list entries.
cursor_pen:Specifies the pen for the selected entry.
inactive_pen:If specified, used for the cursor when the widget is not active.
icon_pen:Default pen for icons.
on_select:Selection change callback; called as on_select(index,choice). This is also called with nil arguments if setChoices is called with an empty list.
on_submit:Enter key callback; if specified, the list reacts to the key and calls it as on_submit(index,choice).
on_submit2:Shift-Enter key callback; if specified, the list reacts to the key and calls it as on_submit2(index,choice).
row_height:Height of every row in text lines.
icon_width:If not nil, the specified number of character columns are reserved to the left of the list item for the icons.
scroll_keys:Specifies which keys the list should react to as a table.

Every list item may be specified either as a string, or as a lua table with the following fields:

text:Specifies the label text in the same format as the Label text.
caption, [1]:Deprecated legacy aliases for text.
text_*:Reserved for internal use.
key:Specifies a keybinding that acts as a shortcut for the specified item.
icon:Specifies an icon string, or a pen to paint a single character. May be a callback.
icon_pen:When the icon is a string, used to paint it.

The list supports the following methods:

  • List{ ..., choices = ..., selected = ... }

    Same as calling setChoices after construction.

  • list:setChoices(choices[, selected])

    Replaces the list of choices, possibly also setting the currently selected index.

  • list:setSelected(selected)

    Sets the currently selected index. Returns the index after validation.

  • list:getChoices()

    Returns the list of choices.

  • list:getSelected()

    Returns the selected index, choice, or nothing if the list is empty.

  • list:getContentWidth()

    Returns the minimal width to draw all choices without clipping.

  • list:getContentHeight()

    Returns the minimal width to draw all choices without scrolling.

  • list:submit()

    Call the on_submit callback, as if the Enter key was handled.

  • list:submit2()

    Call the on_submit2 callback, as if the Shift-Enter key was handled.

FilteredList class

This widget combines List, EditField and Label into a combo-box like construction that allows filtering the list by subwords of its items.

In addition to passing through all attributes supported by List, it supports:

edit_pen:If specified, used instead of cursor_pen for the edit field.
edit_below:If true, the edit field is placed below the list instead of above.
edit_key:If specified, the edit field is disabled until this key is pressed.
 Specifies the text of the label shown when no items match the filter.

The list choices may include the following attributes:

search_key:If specified, used instead of text to match against the filter. This is required for any entries where text is not a string.

The widget implements:

  • list:setChoices(choices[, selected])

    Resets the filter, and passes through to the inner list.

  • list:getChoices()

    Returns the list of all choices.

  • list:getVisibleChoices()

    Returns the filtered list of choices.

  • list:getFilter()

    Returns the current filter string, and the filtered list of choices.

  • list:setFilter(filter[,pos])

    Sets the new filter string, filters the list, and selects the item at index pos in the unfiltered list if possible.

  • list:canSubmit()

    Checks if there are currently any choices in the filtered list.

  • list:getSelected(), list:getContentWidth(), list:getContentHeight(), list:submit()

    Same as with an ordinary list.


DFHack plugins may export native functions and events to Lua contexts. These are exposed as plugins.<name> modules, which can be imported with require('plugins.<name>'). The plugins listed in this section expose functions and/or data to Lua in this way.

In addition to any native functions documented here, plugins that can be enabled (that is, plugins that support the enable/disable API) will have the following functions defined:

  • isEnabled() returns whether the plugin is enabled.
  • setEnabled(boolean) sets whether the plugin is enabled.

For plugin developers, note that a Lua file in plugins/lua is required for require() to work, even if it contains no pure-Lua functions. This file must contain mkmodule('plugins.<name>') to import any native functions defined in the plugin. See existing files in plugins/lua for examples.


Native functions provided by the blueprint plugin:

  • dig(start, end, name)

  • build(start, end, name)

  • place(start, end, name)

  • query(start, end, name)

    start and end are tables containing positions (see xyz2pos). name is used as the basis for the filename.


This plugin overwrites some methods in workshop df class so that mechanical workshops are possible. Although plugin export a function it’s recommended to use lua decorated function.


registerBuilding(table) where table must contain name, as a workshop raw name, the rest are optional:


custom workshop id e.g. SOAPMAKER


this is the only mandatory field.


if true make impassible tiles impassible to liquids too


how much machine power is needed to work. Disables reactions if not supplied enough and needs_power==1


how much machine power is produced.


if produced in network < consumed stop working, default true


a table or {x=?,y=?} of connection points for machines.


a table of number (how much ticks to skip) and a function which gets called on shop update


a table of frames which can be a table of:

  1. tables of 4 numbers {tile,fore,back,bright} OR
  2. empty table (tile not modified) OR
  3. {x=<number> y=<number> + 4 numbers like in first case}, this generates full frame useful for animations that change little (1-2 tiles)

a flag if this building can be counted in room. 1 means it can, 0 means it can’t and -1 default building behaviour


a flag that automatically fills up gears and animate. It looks over building definition for gear icons and maps them.

Animate table also might contain:

frameLength:how many ticks does one frame take OR
isMechanical:a bool that says to try to match to mechanical system (i.e. how gears are turning)

getPower(building) returns two number - produced and consumed power if building can be modified and returns nothing otherwise

setPower(building,produced,consumed) sets current productiona and consumption for a building.


Simple mechanical workshop:

  gears={x=0,y=0}, --connection point
    isMechanical=true, --animate the same conn. point as vanilla gear
    {{x=0,y=0,42,7,0,0}}, --first frame, 1 changed tile
    {{x=0,y=0,15,7,0,0}} -- second frame, same

Or with auto_gears:



Native functions provided by the buildingplan plugin:

  • bool isPlannableBuilding(df::building_type type, int16_t subtype, int32_t custom) returns whether the building type is handled by buildingplan.
  • bool isPlanModeEnabled(df::building_type type, int16_t subtype, int32_t custom) returns whether the buildingplan UI is enabled for the specified building type.
  • bool isPlannedBuilding(df::building *bld) returns whether the given building is managed by buildingplan.
  • void addPlannedBuilding(df::building *bld) suspends the building jobs and adds the building to the monitor list.
  • void doCycle() runs a check for whether buildlings in the monitor list can be assigned items and unsuspended. This method runs automatically twice a game day, so you only need to call it directly if you want buildingplan to do a check right now.
  • void scheduleCycle() schedules a cycle to be run during the next non-paused game frame. Can be called multiple times while the game is paused and only one cycle will be scheduled.


The burrows plugin implements extended burrow manipulations.


  • onBurrowRename.foo = function(burrow)

    Emitted when a burrow might have been renamed either through the game UI, or renameBurrow().

  • onDigComplete.foo = function(job_type,pos,old_tiletype,new_tiletype,worker)

    Emitted when a tile might have been dug out. Only tracked if the auto-growing burrows feature is enabled.

Native functions:

  • renameBurrow(burrow,name)

    Renames the burrow, emitting onBurrowRename and updating auto-grow state properly.

  • findByName(burrow,name)

    Finds a burrow by name, using the same rules as the plugin command line interface. Namely, trailing '+' characters marking auto-grow burrows are ignored.

  • copyUnits(target,source,enable)

    Applies units from source burrow to target. The enable parameter specifies if they are to be added or removed.

  • copyTiles(target,source,enable)

    Applies tiles from source burrow to target. The enable parameter specifies if they are to be added or removed.

  • setTilesByKeyword(target,keyword,enable)

    Adds or removes tiles matching a predefined keyword. The keyword set is the same as used by the command line.

The lua module file also re-exports functions from dfhack.burrows.


Exposes some features of the C++11 random number library to Lua.

Native functions (exported to Lua)

  • GenerateEngine(seed)

    returns engine id

  • DestroyEngine(rngID)

    destroys corresponding engine

  • NewSeed(rngID, seed)

    re-seeds engine

  • rollInt(rngID, min, max)

    generates random integer

  • rollDouble(rngID, min, max)

    generates random double

  • rollNormal(rngID, avg, stddev)

    generates random normal[gaus.]

  • rollBool(rngID, chance)

    generates random boolean

  • MakeNumSequence(start, end)

    returns sequence id

  • AddToSequence(seqID, num)

    adds a number to the sequence

  • ShuffleSequence(rngID, seqID)

    shuffles the number sequence

  • NextInSequence(seqID)

    returns the next number in sequence

Lua plugin functions

  • MakeNewEngine(seed)

    returns engine id

Lua plugin classes

  • init(id, df, dist): constructor
    • id: Reference ID of engine to use in RNGenerations
    • df (optional): bool indicating whether to destroy the Engine when the crng object is garbage collected
    • dist (optional): lua number distribution to use
  • changeSeed(seed): alters engine’s seed value
  • setNumDistrib(distrib): sets the number distribution crng object should use
    • distrib: number distribution object to use in RNGenerations
  • next(): returns the next number in the distribution
  • shuffle(): effectively shuffles the number distribution
  • init(avg, stddev): constructor
  • next(id): returns next number in the distribution
    • id: engine ID to pass to native function
  • init(min, max): constructor
  • next(id): returns next number in the distribution
    • id: engine ID to pass to native function
  • init(min, max): constructor
  • next(id): returns next number in the distribution
    • id: engine ID to pass to native function
  • init(min, max): constructor
  • next(id): returns next boolean in the distribution
    • id: engine ID to pass to native function
  • init(a, b): constructor
  • add(num): adds num to the end of the number sequence
  • shuffle(): shuffles the sequence of numbers
  • next(): returns next number in the sequence


This plugin exports some events to lua thus allowing to run lua functions on DF world events.

List of events

  1. onReactionComplete(reaction,reaction_product,unit,input_items,input_reagents,output_items,call_native)

    Auto activates if detects reactions starting with LUA_HOOK_. Is called when reaction finishes.

  2. onItemContaminateWound(item,unit,wound,number1,number2)

    Is called when item tries to contaminate wound (e.g. stuck in).

  3. onProjItemCheckMovement(projectile)

    Is called when projectile moves.

  4. onProjItemCheckImpact(projectile,somebool)

    Is called when projectile hits something.

  5. onProjUnitCheckMovement(projectile)

    Is called when projectile moves.

  6. onProjUnitCheckImpact(projectile,somebool)

    Is called when projectile hits something.

  7. onWorkshopFillSidebarMenu(workshop,callnative)

    Is called when viewing a workshop in ‘q’ mode, to populate reactions, useful for custom viewscreens for shops.

  8. postWorkshopFillSidebarMenu(workshop)

    Is called after calling (or not) native fillSidebarMenu(). Useful for job button tweaking (e.g. adding custom reactions)

Events from EventManager

These events are straight from EventManager module. Each of them first needs to be enabled. See functions for more info. If you register a listener before the game is loaded, be aware that no events will be triggered immediately after loading, so you might need to add another event listener for when the game first loads in some cases.

  1. onBuildingCreatedDestroyed(building_id)

    Gets called when building is created or destroyed.

  2. onConstructionCreatedDestroyed(building_id)

    Gets called when construction is created or destroyed.

  3. onJobInitiated(job)

    Gets called when job is issued.

  4. onJobCompleted(job)

    Gets called when job is finished. The job that is passed to this function is a copy. Requires a frequency of 0 in order to distinguish between workshop jobs that were cancelled by the user and workshop jobs that completed successfully.

  5. onUnitDeath(unit_id)

    Gets called on unit death.

  6. onItemCreated(item_id)

    Gets called when item is created (except due to traders, migrants, invaders and spider webs).

  7. onSyndrome(unit_id,syndrome_index)

    Gets called when new syndrome appears on a unit.

  8. onInvasion(invasion_id)

    Gets called when new invasion happens.

  9. onInventoryChange(unit_id,item_id,old_equip,new_equip)

    Gets called when someone picks up an item, puts one down, or changes the way they are holding it. If an item is picked up, old_equip will be null. If an item is dropped, new_equip will be null. If an item is re-equipped in a new way, then neither will be null. You absolutely must NOT alter either old_equip or new_equip or you might break other plugins.

  10. onReport(reportId)

    Gets called when a report happens. This happens more often than you probably think, even if it doesn’t show up in the announcements.

  11. onUnitAttack(attackerId, defenderId, woundId)

    Called when a unit wounds another with a weapon. Is NOT called if blocked, dodged, deflected, or parried.

  12. onUnload()

    A convenience event in case you don’t want to register for every onStateChange event.

  13. onInteraction(attackVerb, defendVerb, attackerId, defenderId, attackReportId, defendReportId)

    Called when a unit uses an interaction on another.


  1. registerReaction(reaction_name,callback)

    Simplified way of using onReactionComplete; the callback is function (same params as event).

  2. removeNative(shop_name)

    Removes native choice list from the building.

  3. addReactionToShop(reaction_name,shop_name)

    Add a custom reaction to the building.

  4. enableEvent(evType,frequency)

    Enable event checking for EventManager events. For event types use eventType table. Note that different types of events require different frequencies to be effective. The frequency is how many ticks EventManager will wait before checking if that type of event has happened. If multiple scripts or plugins use the same event type, the smallest frequency is the one that is used, so you might get events triggered more often than the frequency you use here.

  5. registerSidebar(shop_name,callback)

    Enable callback when sidebar for shop_name is drawn. Usefull for custom workshop views e.g. using gui.dwarfmode lib. Also accepts a class instead of function as callback. Best used with gui.dwarfmode class WorkshopOverlay.


Spawn dragon breath on each item attempt to contaminate wound:

b=require "plugins.eventful"
    local flw=dfhack.maps.spawnFlow(unit.pos,6,0,0,50000)

Reaction complete example:

b=require "plugins.eventful"

  local pos=copyall(unit.pos)
  -- spawn dragonbreath after 100 ticks
  dfhack.timeout(100,"ticks",function() dfhack.maps.spawnFlow(pos,6,0,0,50000) end)
  --do not call real item creation code

Grenade example:

b=require "plugins.eventful"
  -- you can check if projectile.item e.g. has correct material

Integrated tannery:

b=require "plugins.eventful"


A way to access csocket from lua. The usage is made similar to luasocket in vanilla lua distributions. Currently only subset of functions exist and only tcp mode is implemented.

Socket class

This is a base class for client and server sockets. You can not create it - it’s like a virtual base class in c++.

  • socket:close()

    Closes the connection.

  • socket:setTimeout(sec,msec)

    Sets the operation timeout for this socket. It’s possible to set timeout to 0. Then it performs like a non-blocking socket.

Client class

Client is a connection socket to a server. You can get this object either from tcp:connect(address,port) or from server:accept(). It’s a subclass of socket.

  • client:receive(pattern)

    Receives data. Pattern is one of:

    *l:read one line (default, if pattern is nil)
    <number>:read specified number of bytes
    *a:read all available data
  • client:send(data)

    Sends data. Data is a string.

Server class

Server is a socket that is waiting for clients. You can get this object from tcp:bind(address,port).

  • server:accept()

    Accepts an incoming connection if it exists. Returns a client object representing that socket.

Tcp class

A class with all the tcp functionality.

  • tcp:bind(address,port)

    Starts listening on that port for incoming connections. Returns server object.

  • tcp:connect(address,port)

    Tries connecting to that address and port. Returns client object.


A way to ask DF to render a section of the fortress mode map. This uses a native DF rendering function so it’s highly dependent on DF settings (e.g. tileset, colors, etc.)


  • render_map_rect(x,y,z,w,h)

    returns a table with w*h*4 entries of rendered tiles. The format is same as df.global.gps.screen (tile,foreground,bright,background).


This plugin implements the back end of the gui/pathable script. It exports a single Lua function, in hack/lua/plugins/pathable.lua:

  • paintScreen(cursor[,skip_unrevealed]): Paint each visible of the screen green or red, depending on whether it can be pathed to from the tile at cursor. If skip_unrevealed is specified and true, do not draw unrevealed tiles.


The sort plugin does not export any native functions as of now. Instead, it calls Lua code to perform the actual ordering of list items.


Utility functions to facilitate reading .xlsx spreadsheets. It provides the following low-level API methods:

  • open_xlsx_file(filename) returns a file_handle or nil on error
  • close_xlsx_file(file_handle) closes the specified file_handle
  • list_sheets(file_handle) returns a list of strings representing sheet names
  • open_sheet(file_handle, sheet_name) returns a sheet_handle. This call always succeeds, even if the sheet doesn’t exist. Non-existent sheets will have no data, though.
  • close_sheet(sheet_handle) closes the specified sheet_handle
  • get_row(sheet_handle, max_tokens) returns a list of strings representing the contents of the cells in the next row. The max_tokens parameter is optional. If set to a number > 0, it limits the number of cells read and returned for the row.

The plugin also provides Lua class wrappers for ease of use:

  • XlsxioReader provides access to .xlsx files
  • XlsxioSheetReader provides access to sheets within .xlsx files
  • open(filepath) initializes and returns an XlsxioReader object

The XlsxioReader class has the following methods:

  • XlsxioReader:close() closes the file. Be sure to close any open child sheet handles first!
  • XlsxioReader:list_sheets() returns a list of strings representing sheet names
  • XlsxioReader:open_sheet(sheet_name) returns an initialized XlsxioSheetReader object

The XlsxioSheetReader class has the following methods:

  • XlsxioSheetReader:close() closes the sheet
  • XlsxioSheetReader:get_row(max_tokens) reads the next row from the sheet. If max_tokens is specified and is a positive integer, only the first max_tokens elements of the row are returned.

Here is an end-to-end example:

local xlsxreader = require('plugins.xlsxreader')

local function dump_sheet(reader, sheet_name)
    print('reading sheet: ' .. sheet_name)
    local sheet_reader = reader:open_sheet(sheet_name)
        function() sheet_reader:close() end,
            local row_cells = sheet_reader:get_row()
            while row_cells do
                row_cells = sheet_reader:get_row()

local filepath = 'path/to/some_file.xlsx'
local reader = xlsxreader.open(filepath)
    function() reader:close() end,
        for _,sheet_name in ipairs(reader:list_sheets()) do
            dump_sheet(reader, sheet_name)


Any files with the .lua extension placed into the hack/scripts folder are automatically made avaiable as DFHack commands. The command corresponding to a script is simply the script’s filename, relative to the scripts folder, with the extension omitted. For example:

  • hack/scripts/add-thought.lua is invoked as add-thought
  • hack/scripts/gui/teleport.lua is invoked as gui/teleport


Scripts placed in subdirectories can be run as described above, but are not listed by the ls command unless -a is specified. In general, scripts should be placed in subfolders in the following situations:

  • devel: scripts that are intended exclusively for DFHack development, including examples, or scripts that are experimental and unstable
  • fix: fixes for specific DF issues
  • gui: GUI front-ends for existing tools (for example, see the relationship between teleport and gui/teleport)
  • modtools: scripts that are intended to be run exclusively as part of mods, not directly by end-users (as a rule of thumb: if someone other than a mod developer would want to run a script from the console, it should not be placed in this folder)

Scripts can also be placed in other folders - by default, these include raw/scripts and data/save/region/raw/scripts, but additional folders can be added (for example, a copy of the scripts repository for local development). See Script paths for more information on how to configure this behavior.

If the first line of the script is a one-line comment (starting with --), the content of the comment is used by the built-in ls and help commands. Such a comment is required for every script in the official DFHack repository.

Scripts are read from disk when run for the first time, or if they have changed since the last time they were run.

Each script has an isolated environment where global variables set by the script are stored. Values of globals persist across script runs in the same DF session. See devel/lua-example for an example of this behavior. Note that local variables do not persist.

Arguments are passed in to the scripts via the ... built-in quasi-variable; when the script is called by the DFHack core, they are all guaranteed to be non-nil strings.

Additional data about how a script is invoked is passed to the script as a special dfhack_flags global, which is unique to each script. This table is guaranteed to exist, but individual entries may be present or absent depending on how the script was invoked. Flags that are present are described in the subsections below.

DFHack invokes the scripts in the core context; however it is possible to call them from any lua code (including from other scripts) in any context with dfhack.run_script() below.

General script API

  • dfhack.run_script(name[,args...])

    Run a Lua script in hack/scripts/, as if it were started from the DFHack command-line. The name argument should be the name of the script without its extension, as it would be used on the command line.


    In DFHack prompt:

    repeat -time 14 -timeUnits days -command [ workorder ShearCreature ] -name autoShearCreature

    In Lua script:

    dfhack.run_script("repeat", "-time", "14", "-timeUnits", "days", "-command", "[", "workorder", "ShearCreature", "]", "-name", "autoShearCreature")

    Note that the dfhack.run_script() function allows Lua errors to propagate to the caller.

    To run other types of commands (such as built-in commands, plugin commands, or Ruby scripts), see dfhack.run_command(). Note that this is slightly slower than dfhack.run_script() for Lua scripts.

  • dfhack.script_help([name, [extension]])

    Returns the contents of the embedded documentation of the specified script. extension defaults to “lua”, and name defaults to the name of the script where this function was called. For example, the following can be used to print the current script’s help text:

    local args = {...}
    if args[1] == 'help' then

Importing scripts

  • dfhack.reqscript(name) or reqscript(name)

    Loads a Lua script and returns its environment (i.e. a table of all global functions and variables). This is similar to the built-in require(), but searches all script paths for the first matching name.lua file instead of searching the Lua library paths (like hack/lua).

    Most scripts can be made to support reqscript() without significant changes (in contrast, require() requires the use of mkmodule() and some additional boilerplate). However, because scripts can have side effects when they are loaded (such as printing messages or modifying the game state), scripts that intend to support being imported must satisfy some criteria to ensure that they can be imported safely:

    1. Include the following line - reqscript() will fail if this line is not present:

      --@ module = true
    2. Include a check for dfhack_flags.module, and avoid running any code that has side-effects if this flag is true. For instance:

      -- (function definitions)
      if dfhack_flags.module then
      -- (main script code with side-effects)


      -- (function definitions)
      function main()
          -- (main script code with side-effects)
      if not dfhack_flags.module then

    Example usage:

    local addThought = reqscript('add-thought')
    addThought.addEmotionToUnit(unit, ...)

    Circular dependencies between scripts are supported, as long as the scripts have no side-effects at load time (which should already be the case per the above criteria).


    Avoid caching the table returned by reqscript() beyond storing it in a local or global variable as in the example above. reqscript() is fast for scripts that have previously been loaded and haven’t changed. If you retain a reference to a table returned by an old reqscript() call, this may lead to unintended behavior if the location of the script changes (e.g. if a save is loaded or unloaded, or if a script path is added in some other way).


    Mods that include custom Lua modules can write these modules to support reqscript() and distribute them as scripts in raw/scripts. Since the entire raw folder is copied into new saves, this will allow saves to be successfully transferred to other users who do not have the mod installed (as long as they have DFHack installed).

    Backwards compatibility notes

    For backwards compatibility, moduleMode is also defined if dfhack_flags.module is defined, and is set to the same value. Support for this may be removed in a future version.

  • dfhack.script_environment(name)

    Similar to reqscript() but does not enforce the check for module support. This can be used to import scripts that support being used as a module but do not declare support as described above, although it is preferred to update such scripts so that reqscript() can be used instead.

Enabling and disabling scripts

Scripts can choose to recognize the built-in enable and disable commands by including the following line anywhere in their file:

--@ enable = true

When the enable and disable commands are invoked, the dfhack_flags table passed to the script will have the following fields set:

  • enable: Always true if the script is being enabled or disabled
  • enable_state: true if the script is being enabled, false otherwise

Example usage:

--@ enable = true
-- (function definitions...)
if dfhack_flags.enable then
    if dfhack_flags.enable_state then

Save init script

If a save directory contains a file called raw/init.lua, it is automatically loaded and executed every time the save is loaded. The same applies to any files called raw/init.d/*.lua. Every such script can define the following functions to be called by dfhack:

  • function onStateChange(op) ... end

    Automatically called from the regular onStateChange event as long as the save is still loaded. This avoids the need to install a hook into the global dfhack.onStateChange table, with associated cleanup concerns.

  • function onUnload() ... end

    Called when the save containing the script is unloaded. This function should clean up any global hooks installed by the script. Note that when this is called, the world is already completely unloaded.

Within the init script, the path to the save directory is available as SAVE_PATH.