local glue = require'glue'

API Summary

Math

glue.round(x[, p]) -> y
glue.snap(x[, p]) -> y

Round a number towards nearest integer or multiple of p. Implemented as math.floor(x / p + .5) * p. Rounds half-up (i.e. it returns -1 for -1.5). Works with numbers up to +/-2^52. It’s not dead accurate as it returns eg. 1 instead of 0 for 0.49999999999999997 (the number right before 0.5) which is < 0.5.

glue.floor(x[, p]) -> y

Round a number towards nearest smaller integer or multiple of p. Implemented as math.floor(x / p) * p.

glue.ceil(x[, p]) -> y

Round a number towards nearest larger integer or multiple of p. Implemented as math.ceil(x / p) * p.

glue.clamp(x, min, max)

Clamp a value in range. Implemented as math.min(math.max(x, min), max), so if max < min, the result is max.

glue.lerp(x, x0, x1, y0, y1) -> y

Linear interpolation, i.e. linearly project x in x0..x1 range to the y0..y1 range.

glue.sign(x) -> 1|0|-1

Return sign of x.

glue.strict_sign(x) -> 1|-1

Return strict sign of x.

glue.nextpow2(x) -> y

Find the smallest n for which x <= 2^n.

glue.repl(x, v, r) -> x

If x == v, return r, otherwise return x.

glue.random_string(n) -> s

Generate random string of length n.

glue.uuid() -> s

Generate random UUID (v4).

Don’t forget to seed the randomizer first, eg. with math.randomseed(require'time'.clock()) or what have you.

Varargs

glue.pack(...) -> t

Pack varargs. Implemented as n = select('#', ...), ...}.

glue.unpack(t,[i][,j]) -> ...

Unpack varargs. Implemented as unpack(t, i or 1, j or t.n or #t).

Tables

glue.count(t[, maxn]) -> n

Count the keys in a table, optionally up to maxn.

glue.index(t) -> dt

Switch table keys with values.

Examples

Extract a rfc850 date from a string. Use lookup tables for weekdays and months.

local weekdays = glue.index{'Sunday', 'Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday'}
local months = glue.index{'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec'}

--weekday "," SP 2DIGIT "-" month "-" 2DIGIT SP 2DIGIT ":" 2DIGIT ":" 2DIGIT SP "GMT"
--eg. Sunday, 06-Nov-94 08:49:37 GMT
function rfc850date(s)
   local w,d,mo,y,h,m,s = s:match'([A-Za-z]+), (%d+)%-([A-Za-z]+)%-(%d+) (%d+):(%d+):(%d+) GMT'
   d,y,h,m,s = tonumber(d),tonumber(y),tonumber(h),tonumber(m),tonumber(s)
   w = assert(weekdays[w])
   mo = assert(months[mo])
   if y then y = y + (y > 50 and 1900 or 2000) end
   return {wday = w, day = d, year = y, month = mo, hour = h, min = m, sec = s}
end

for k,v in pairs(rfc850date'Sunday, 06-Nov-94 08:49:37 GMT') do
   print(k,v)
end

Output

day   6
sec   37
wday  1
min   49
year  1994
month 11
hour  8

Copy-paste a bunch of defines from a C header file and create an inverse lookup table to find the name of a value at runtime.

--from ibase.h
info_end_codes = {
   isc_info_end             = 1,  --normal ending
   isc_info_truncated       = 2,  --receiving buffer too small
   isc_info_error           = 3,  --error, check status vector
   isc_info_data_not_ready  = 4,  --data not available for some reason
   isc_info_svc_timeout     = 64, --timeout expired
}
info_end_code_names = glue.index(info_end_codes)
print(info_end_code_names[64])

Output

isc_info_svc_timeout

glue.keys(t[,sorted|cmp]) -> dt
glue.sortedkeys(t[,cmp]) -> dt

Make an array of all the keys of t, optionally sorted. The second arg can be true, 'asc', 'desc' or a comparison function.

Examples

An API expects an array of things but you have them as keys in a table because you are indexing something on them.

For instance, you have a table of the form {socket = thread} but socket.select wants an array of sockets.

glue.sortedpairs(t[,cmp]) -> iter() -> k,v

Like pairs() but in key order.

The implementation creates a temporary table to sort the keys in.

glue.update(dt,t1,...) -> dt

Update a table with elements of other tables, overwriting any existing keys.

• falsey arguments are skipped.

Examples

Create an options table by merging the options received as an argument (if any) over the default options.

function f(opts)
   opts = glue.update({}, default_opts, opts)
end

Shallow table copy:

t = glue.update({}, t)

Static multiple inheritance:

C = glue.update({}, A, B)

glue.merge(dt,t1,...) -> dt

Update a table with elements of other tables skipping on any existing keys.

• falsey arguments are skipped.

Examples

Normalize a data object with default values:

glue.merge(t, defaults)

glue.attr(t,k1[,v])[k2] = v

Idiom for t[k1][k2] = v with auto-creating of t[k1] if not present.

Arrays

glue.extend(dt,t1,...) -> dt

Extend an array with the elements of other arrays.

• falsey arguments are skipped.
• array elements are the ones from 1 to #dt.

Uses

Accumulating values from multiple array sources.

glue.append(dt,v1,...) -> dt

Append non-nil arguments to an array.

Uses

Appending an object to a flattened array of arrays (eg. appending a path element to a 2d path).

glue.shift(t,i,n) -> t

Shift all the array elements starting at index i, n positions to the left or further to the right.

For a positive n, shift the elements further to the right, effectively creating room for n new elements at index i. When n is 1, the effect is the same as for table.insert(t, i, t[i]). The old values at index i to i+n-1 are preserved, so #t still works after the shifting.

For a negative n, shift the elements to the left, effectively removing n elements at index i. When n is -1, the effect is the same as for table.remove(t, i).

Uses

Removing a portion of an array or making room for more elements inside the array.

glue.map(t, field|f,...) -> t

Map function f(k, v, ...) -> v1 over the key-value pairs of t or.

If f is not a function, then the values of t must be themselves tables, in which case f is a key to pluck from those tables. Plucked functions are called as methods and their result is selected instead. This allows eg. calling a method for each element in a table of objects and collecting the results in a table.

glue.imap(t, field|f,...) -> t

Map function f(v, ...) -> v1 over the array elements of t taken to be from 1 up to t.n or #t.

If f is not a function, then the values of t must be themselves tables, in which case f is a key to pluck from those tables. Plucked functions are called as methods and their result is selected instead. This allows eg. calling a method for each element in an array of objects and collecting the results in an array.

glue.indexof(v, t, [i], [j]) -> i

Scan an array for a value and if found, return the index.

NOTE: Works on ffi arrays too if i and j are provided.

glue.binsearch(v, t, [cmp], [i], [j]) -> i

Return the smallest index whereby inserting the value v in sorted array t will keep t sorted i.e. t[i-1] < v and t[i] >= v. Return nil if v is larger than the largest value or if t is empty.

The comparison function cmp is called as cmp(t, i, v) and must return true when t[i] < v. Built-in functions are also available by passing one of '<', '>', '<=', '>='.

TIP: Use a cmp that returns true when t[i] > v to search in a reverse-sorted array (i.e. use '>').

TIP: Use a cmp that returns true when t[i] <= v to get the largest index (as opposed to the smallest index) that will keep t sorted when inserting v, i.e. t[i-1] <= v and t[i] > v.

NOTE: Works on ffi arrays too if i and j are provided.

glue.sortedarray([sa]) -> sa

Creates an array that stays sorted with insertion, searching and removal in O(log n) leveraging binary search.

• if given an existing sa to be wrapped, it must be already sorted.
• sa.cmp is used for cmp in binarysearch().
• use sa:push(v) to add values.
• use sa:find(v) -> i|nil to look up values.
• use sa:remove_value(v) -> v|nil to find and remove a value.

glue.reverse(t, [i], [j]) -> t

Reverse an array in-place and return the input arg.

NOTE: Works on ffi arrays too if i and j are provided.

Strings

glue.gsplit(s,sep[,start[,plain]]) -> iter() -> e[,captures...]

glue.split(s,sep[,start[,plain]]) -> {s1,...}

Split a string by a separator pattern (or plain string) and iterate over the elements.

• if sep is "" return the entire string in one iteration
• if s is "" return s in one iteration
• empty strings between separators are always returned, eg. glue.gsplit(',', ',') produces 2 empty strings
• captures are allowed in sep and they are returned after the element, except for the last element for which they don’t match (by definition).

Examples

for s in glue.gsplit('Spam eggs spam spam and ham', '%s*spam%s*') do
   print('"'..s..'"')
end

> "Spam eggs"
> ""
> "and ham"

glue.names('name1 ...') -> {'name1', ...}

Split a string by whitespace. Unlike glue.split(s, '%s+'), it ignores resulting empty elements. Also, non-string args pass through.

glue.capitalize(s) -> s

Capitalize the first letter of every word in string.

glue.lines(s, [opt], [init]) -> iter() -> s, i, j, k

Iterate the lines of a string. For each line it returns the line contents, the content-start, content-end and the next-content-start indices.

• the lines are split at \r\n, \r and \n markers.
• the line ending markers are included or excluded depending on the second arg, which can be *L (include line endings; default) or *l (exclude).
• if the string is empty or doesn’t contain a line ending marker, it is iterated once.
• if the string ends with a line ending marker, one more empty string is iterated.
• init tells it where to start parsing (default is 1).

glue.outdent(s, [indent]) -> s, indent

Remove spaces/tabs indentation of multi-line text based on the indentation of the first line. If a subsequent line is less indented than the first line, returns the original string. If indent given, it is prepended to each line.

glue.lineinfo(s, [i]) -> line, col

Given a byte position in a text, return the text position. If i is not given, returns a function f(i) that is faster on repeat calls.

glue.trim(s) -> s

Remove whitespace (defined as Lua pattern "%s") from the beginning and end of a string.

glue.pad(s, n, [c], dir) -> s
glue.lpad(s, n, [c]) -> s
glue.rpad(s, n, [c]) -> s

Pad a string s to length n using char c (which defaults to ' ') on its right (dir = ‘r’) or left (dir = ‘l’).

glue.esc(s[,mode]) -> pat

Escape magic characters of the string s so that it can be used as a pattern to string matching functions.

• the optional argument mode can have the value "*i" (for case insensitive), in which case each alphabetical character in s will also be escaped as "[aA]" so that it matches both its lowercase and uppercase variants.
• escapes embedded zeroes as the %z pattern.

Uses

• workaround for lack of pattern syntax for “this part of a match is an arbitrary string”
• workaround for lack of a case-insensitive flag in pattern matching functions

glue.tohex(s|n[,upper]) -> s

Convert a binary string or a Lua number to its hex representation.

• lowercase by default
• uppercase if the arg upper is truthy
• numbers must be in the unsigned 32 bit integer range

glue.fromhex(s[, isvalid]) -> s

Convert a hex string to its binary representation. Returns nil on invalid input unless isvalid is true which makes it raise on invalid input.

glue.starts(s, prefix) -> t|f

Find if string s starts with prefix. Implemented as s:sub(1, #p) == p which is 5x faster than s:find'^...' in LuaJIT 2.1 with JIT on (and about the same with jit off).

glue.ends(s, suffix) -> t|f

Find if string s ends with suffix.

glue.subst(s, t) -> s

Replace all {foo} occurences within s with t.foo.

glue.catargs(sep, ...) -> s

Concat args, skipping nil ones. Returns nil on zero non-nil args.

Iterators

glue.collect([i,]iterator) -> t

Iterate an iterator and collect its i’th return value of every step into an array.

• i defaults to 1

Examples

Implementation of keys() and values() in terms of collect()

keys = function(t) return glue.collect(pairs(t)) end
values = function(t) return glue.collect(2,pairs(t)) end

Collecting string matches:

s = 'a,b,c,'
t = glue.collect(s:gmatch'(.-),')
for i=1,#t do print(t[i]) end

> a
> b
> c

Stubs

glue.pass(...) -> ...

The identity function. Does nothing, returns back all arguments.

Uses

Default value for optional callback arguments:

function urlopen(url, callback, errback)
   callback = callback or glue.pass
   errback = errback or glue.pass
   ...
   callback()
end

glue.noop()

Does nothing. Returns nothing.

Caching

glue.memoize(f, [narg], [weak]) -> f

glue.memoize_multiret(f, [narg], [weak]) -> f

Memoization for functions with any number of arguments. memoize() supports functions with one return value. memoize_multiret() supports any function. Both support nil and NaN args and retvals.

Memoization guarantees to only call the original function once for the same combination of arguments.

Special attention is given to the vararg part of the function, if any. For instance, for a function f(x, y, ...), calling f(1) is considered to be the same as calling f(1, nil), but calling f(1, nil) is not the same as calling f(1, nil, nil).

The optional narg argument fixates the function to always take exactly narg args regardless of how the function was defined.

The optional weak argument makes the cache of returned values weak and is useful for caching objects that are pinned elsewere without leaking memory. Using this flag requires that the function to be memoized returns heap objects only and always!

glue.tuples([narg], [weak]) -> f(...) -> t

glue.weaktuples([narg]) -> f(...) -> t

Create a tuple space, which is a function that returns the same identity t for the same list of arguments. It is implemented as:

local tuple_mt = {__call = glue.unpack}
function glue.tuples(...)
   return glue.memoize(function(...)
      return setmetatable(glue.pack(...), tuple_mt)
   end, ...)
end

Tuples are immutable lists that can be used as table keys because they have value semantics since the tuple constructor returns the same identity for the exact same list of identities.

The result tuple can be expanded back by calling it: t() -> args....

NOTE: Tuple elements are indexed internally with a hash tree. Creating a tuple thus takes N hash lookups and M table creations, where N+M is the number of elements in the tuple. Lookup time depends on how dense the tree is on the search path, which depends on how many existing tuples share a first sequence of elements with the tuple being created. In particular, creating tuples out of all permutations of a certain set of values hits the worst case for lookup time, but creates the minimum amount of tables relative to the number of tuples.

glue.tuple([narg]) -> t

Create a tuple in a default global weak tuple space.

Objects

glue.inherit(t, parent) -> t
glue.inherit(t, nil) -> t

Set a table to inherit attributes from a parent table, or clear inheritance.

If the table has no metatable and inheritance has to be set, not cleared, then make it one.

Examples

Logging mixin:

AbstractLogger = glue.inherit({}, function(t,k) error('abstract '..k) end)
NullLogger = glue.inherit({log = function() end}, AbstractLogger)
PrintLogger = glue.inherit({log = function(self,...) print(...) end}, AbstractLogger)

HttpRequest = glue.inherit({
   perform = function(self, url)
      self:log('Requesting', url, '...')
      ...
   end
}, NullLogger)

LoggedRequest = glue.inherit({log = PrintLogger.log}, HttpRequest)

LoggedRequest:perform'http://lua.org/'

> Requesting   http://lua.org/   ...

Defining a module in Lua 5.2

_ENV = glue.inherit({},_G)
...

To get the effect of static (single or multiple) inheritance, use glue.update.

When setting inheritance, you can pass in a function.

Unlike glue.object, this doesn’t add any keys to the object.

glue.object([super][, t], ...) -> t

Create a class or object from t (which defaults to {}) by setting t as its own metatable, setting t.__index to super and t.__call to super.__call. Extra args are passed to glue.update(self, ...). This simple object model has the following qualities:

• the implementation is only 4 LOC (14 LOC if extra args are used) and can thus be copy-pasted into any module to avoid a dependency on the glue library.
• funcall-style instantiation with t(...) which calls t:__call(...).
• small memory footprint (3 table slots and no additional tables).
• subclassing from instances is allowed (prototype-based inheritance).
• glue.object can serve as a stub class/instance constructor: t.__call = glue.object (t.new = glue.object works too).
• a separate constructor to be used only for subclassing can be made with the same pattern: t.subclass = glue.object.
• virtual classes (aka dependency injection, aka nested inner classes whose fields and methods can be overridden by subclasses of the outer class): composite objects which need to instantiate other objects can be made extensible by exposing those objects’ classes as fields of the container class with container_class.inner_class = inner_class and instantiating with self.inner_class(...) so that replacing inner_class in a sub-class of container_class is possible. Moreso, instantiation with self:inner_class(...) (so with a colon) passes the container object to inner_class’s constructor automatically which allows referencing the container object from the inner object.
• overriding syntax sugar so that the super class need not be referenced explicitly when overriding can be incorporated into the base class with base.override = glue.override.

glue.before(class, method_name, f)

Modify a method such that it calls f at the beginning. f receives all the arguments passed to the method. f’s results are discarded.

Usage:

glue.before(foo, 'bar', function(self, ...)
    ...
end)

Alternatively,

foo.before = glue.before
foo:before('bar', function(self, ...)
  ...
end)

glue.after(class, method_name, f)

Modify a method such that it calls f at the end. f receives all the arguments passed to the method. The modified method returns what f returns.

Usage:

glue.after(foo, 'bar', function(self, ...)
    ...
end)

Alternatively,

foo.after = glue.after
foo:after('bar', function(self, ...)
  ...
end)

glue.override(class, method_name, f)

Override a method such that the new implementation only calls f as f(inherited, self, ...) where inherited is the old implementation. f receives all the method arguments and the method returns what f returns.

Usage:

glue.override(foo, 'bar', function(inherited, self, ...)
  ...
  local ret = inherited(self, ...)
  ...
end)

Alternatively,

foo.override = glue.override
foo:override('bar', function(inherited, self, ...)
  ...
  local ret = inherited(self, ...)
  ...
end)

glue.gettersandsetters([getters], [setters], [super]) -> mt

Return a metatable that supports virtual properties with getters and setters. Can be used with setmetatable() and ffi.metatype(). super is for preserving the functionality of __index while __index is being used for getters.

I/O

glue.canopen(file[, mode]) -> filename | nil

Checks whether a file exists and it’s available for reading or writing. The mode arg is the same as for [io.open] and defaults to ‘rb’.

glue.readfile(filename[,format][,open]) -> s | nil, err

Read the contents of a file into a string.

• format can be "t" in which case the file will be read in text mode (default is binary mode).
• open is the file open function which defaults to io.open.

glue.readpipe(cmd[,format][,open]) -> s | nil, err

Read the output of a command into a string. The options are the same as for glue.readfile.

glue.replacefile(oldpath, newpath)

Move or rename a file. If newpath exists and it’s a file, it is replaced by the old file atomically. The operation can still fail under many circumstances like if newpath is a directory or if the files are in different filesystems or if oldpath is missing or locked, etc.

For consistent behavior across OSes, both paths should be absolute paths or just filenames.

On LuaJIT, this is implemented based on MoveFileExA on Windows.

glue.writefile(filename,s|t|read,[format],[tmpfile]) -> ok, err

Write the contents of a string, table or iterator to a file.

• the contents can be given as a string, an array of strings, or a function that returns a string or nil to signal end-of-stream.
• format can be "t" in which case the file will be written in text mode (default is binary mode). It can also be "a" or "at" for appending.
• tmpfile enables atomic saving via a temporary file (enabled by default) which is then renamed to filename and if writing or renaming fails the temp file is removed and filename is not touched (and if the program is killed while writing, you get a stale temp file but no data corruption). If tmpfile is false and writing fails then filename is removed (and if the program is killed while writing, you get a partially written file).

glue.printer(out[, format]) -> f

Create a print()-like function which uses the function out to output its values and uses the optional format to format each value. For instance glue.printer(io.write, tostring) returns a function which behaves like the standard print() function.

glue.time([utc, ][t]) -> ts
glue.time([utc, ][year, [month], [day], [hour], [min], [sec], [isdst]]) -> ts

Like os.time() but considers the arguments to be in UTC if either utc or t.utc is true.

NOTE: You should only use os.date() and os.time() and therefore glue.time() for current dates and use something else for historical dates because these functions don’t work with negative timestamps because apparently time didn’t exist before UNIX. At least they don’t suffer from Y2038 so that’s that.

NOTE: os.time() has second accuracy (so those timestamps are integers). For sub-second accuracy use the time module.

glue.utc_diff([t]) -> seconds

Difference between local time and UTC in seconds.

glue.day([utc, ][ts], [plus_days]) -> ts

Timestamp at day’s beginning from ts, plus/minus some days.

glue.month([utc, ][ts], [plus_months]) -> ts

Timestamp at month’s beginning from ts, plus/minus some months.

glue.year([utc, ][ts], [plus_years]) -> ts

Timestamp at year’s beginning from ts, plus/minus some years.

glue.timeago(ts[, from_ts]) -> s

Format relative time, eg. 3 hours ago or in 2 weeks.

Errors

glue.assert(v[,message[,format_args...]])

Like assert but supports formatting of the error message using string.format().

This is better than assert(v, string.format(message, format_args...)) because it avoids creating the message string when the assertion is true.

CAVEAT: Unlike standard assert(), this only returns its first argument even when no message is given, to avoid returning the error message and its args when a message is given and the assertion is true. So the pattern a, b = glue.assert(f()) doesn’t work.

Example

glue.assert(depth <= maxdepth, 'maximum depth %d exceeded', maxdepth)

glue.protect(func) -> protected_func

In Lua, API functions conventionally signal errors by returning nil and an error message instead of raising errors. In the implementation however, using assert() and error() is preferred to coding explicit conditional flows to cover exceptional cases. Use this function to convert error-raising functions to nil,err-returning functions:

protected_function = glue.protect(function()
   ...
   assert(...)
   ...
   error(...)
   ...
   return result_value
end)

local ret, err = protected_function()

glue.pcall(f,...) -> true,... | false,traceback

With Lua’s pcall() you lose the stack trace, and with usual uses of pcall() you don’t want that. This variant appends the traceback to the error message.

NOTE: Lua 5.2 and LuaJIT only.

glue.fpcall(f,...) -> result | nil,traceback

glue.fcall(f,...) -> result

These constructs bring the try/finally/except idiom to Lua. The first variant returns nil,error when errors occur while the second re-raises the error.

Example

local result = glue.fpcall(function(finally, except, ...)
  local temporary_resource = acquire_resource()
  finally(function() temporary_resource:free() end)
  ...
  local final_resource = acquire_resource()
  except(function() final_resource:free() end)
  ... code that might break ...
  return final_resource
end, ...)

NOTE: Lua 5.2 and LuaJIT only.

Modules

glue.module([name, ][parent]) -> M, P

glue.module([parent, ][name]) -> M, P

Create a module with a public and private namespace and set the environment of the calling function (not the global one!) to the module’s private namespace and return the namespaces. Cross-references between the namespaces are also created at M._P, P._M, P._P and M._M, so both _P and _M can be accessed directly from the new environment.

parent controls what the namespaces will inherit and it can be either another module, in which case M inherits parent and P inherits parent._P, or it can be a string in which case the module to inherit is first required. parent defaults to _M so that calling glue.module() creates a submodule of the current module. If there’s no _M in the current environment then P inherits _G and M inherits nothing.

Specifying a name for the module either returns package.loaded[name] if it is set or creates a module, sets package.loaded[name] to it and returns that. This is useful for creating and referencing shared namespaces without having to make a Lua file and require that.

Naming the module also sets P[name] = M so that public symbols can be declared in foo.bar style instead of _M.bar.

Setting foo.module = glue.module makes module foo directly extensible by calling foo:module'bar' or require'foo':module'bar'.

NOTE: All that functionality is done in just 27 LOC, less than it takes to explain it, so read the code to get to another level of clarity.

glue.autoload(t, submodules) -> t
glue.autoload(t, key, module|loader) -> t

Assign a metatable to t (or override an existing metatable’s __index) such that when a missing key is accessed, the module said to contain that key is require’d automatically.

The submodules argument is table of form {key = module_name | load_function} specifying the corresponding Lua module (or load function) that make each key available to t. The alternative syntax allows specifying the key - submodule associations one by one.

Motivation

Module autoloading allows splitting the implementation of a module in many submodules containing optional, self-contained functionality, without having to make this visible in the user API. This effectively disconnects how an API is modularized from how its implementation is modularized, allowing the implementation to be refactored at a later time without changing the API.

Example

main module (foo.lua):

local function bar() --function implemented in the main module
  ...
end

--create and return the module table
return glue.autoload({
   ...
   bar = bar,
}, {
   baz = 'foo_extra', --autoloaded function, implemented in module foo_extra
})

submodule (foo_extra.lua):

local foo = require'foo'

function foo.baz(...)
  ...
end

in usage:

local foo = require'foo'

foo.baz(...) -- foo_extra was now loaded automatically

glue.bin

Get the script’s directory. This allows finding files in the script’s directory regardless of the directory that Lua is started in.

For executables created with bundle, this is the executable’s directory.

Example

local foobar = glue.readfile(glue.bin .. '/' .. file_near_this_script)

Caveats

This only works if glue itself can already be found and required (chicken/egg problem). Also, the path is relative to the current directory, so this stops working as soon as the current directory is changed. Also, depending on how the process was started, this information might be missing or wrong since it’s set by the parent process. Better use fs.exedir which has none of these problems.

glue.luapath(path[,index[,ext]])

Insert a Lua search pattern in package.path such that require will be able to load Lua modules from that path. The optional index arg specifies the insert position (default is 1, that is, before all existing paths; can be negative, to start counting from the end; can be the string ‘after’, which is the same as 0). The optional ext arg specifies the file extension to use (default is “lua”).

glue.cpath(path[,index])

Insert a Lua search pattern in package.cpath such that require will be able to load Lua/C modules from that path. The index arg has the same meaning as with glue.luapath.

Example

glue.luapath(glue.bin)
glue.cpath(glue.bin)

require'foo' --looking for `foo` in the same directory as the running script first

Allocation

glue.freelist([create], [destroy]) -> alloc, free

Returns alloc() -> e and free(e) functions to allocate and deallocate Lua objects. The allocator returns the last freed object or calls create() to create a new one if the freelist is empty. create defaults to function() return {} end; destroy defaults to glue.noop.

glue.buffer(ctype) -> alloc(minlen|false) -> buf, capacity

(LuaJIT only) Return an allocation function that reuses or reallocates an internal buffer based on the len argument.

• ctype must be a VLA: the returned buffer will have that type. this makes glue.buffer(ctype) compatible with ffi.typeof(ctype).
• the buffer only grows, it never shrinks and it only grows in powers of two steps.
• the allocation function returns the buffer’s current capacity which can be equal or greater than the requested length.
• the returned buffer is anchored by the allocation function. calling alloc(false) unanchors the buffer.
• the contents of the buffer are not preserved between allocations but you are allowed to access both buffers between two consecutive allocations in order to do that yourself.

glue.dynarray(ctype[, min_capacity]) -> alloc(minlen|false) -> buf, minlen

Like glue.buffer() but preserves data between reallocations, and always returns minlen instead of capacity.

glue.readall(read, self, ...) -> buf, len

Repeat read based on a read(self, buf, len, ...) -> readlen function.

glue.buffer_reader(buf,len | nil,err) -> read

Return a read(buf, len) -> readlen function that consumes data from the supplied buffer. The supplied buf,len can also be nil,err in which case the read function will always return just that. The buffer must be a (u)int8_t pointer or VLA.

FFI

glue.addr(ptr) -> number | string

Convert the address of a pointer into a Lua number (or possibly string on 64bit platforms). This is useful for:

• hashing on pointer values (i.e. using pointers as table keys)
• moving pointers in and out of Lua states when using luastate

glue.ptr([ctype,]number|string) -> ptr

Convert an address value stored as a Lua number or string to a cdata pointer, optionally specifying a ctype for the pointer (defaults to void*).

glue.getbit(val, mask) -> true|false

Get the value of a single bit from an integer.

glue.setbit(val, mask, bitval) -> val

Set the value of a single bit from an integer.

glue.bor(flags, bits, [strict]) -> mask

bit.bor() that takes its arguments as a string of form 'opt1 opt2 ...', a list of form {'opt1', 'opt2', ...} or a map of form {opt->true} and performs bit.bor() on the numeric values of those arguments where the numeric values are given as the bits table of form {opt->bitvalue}.

Useful for Luaizing C functions that take bitmask flags.

Example: glue.bor('a c', {a=1, b=2, c=4}) -> 5.

Tips

String functions are also in the glue.string table. You can extend the Lua string namespace:

`glue.update(string, glue.string)`

so you can use them as string methods:

`s:trim()`

Design

glue_design

Last updated: 17 months ago | Edit on GitHub