This is a mirror page, please see the original page:

https://xmake.io/#/manual/extension_modules

All expansion modules need to be imported through the import interface.

core.base.option

Commonly used to get the value of the xmake command parameter option, often used for plugin development.

option.get

Used to get parameter option values in plugin development, for example:

-- import option module
import("core.base.option")

-- plugin entry function
function main(...)
    print(option.get("info"))
end

The above code gets the hello plugin and executes: xmake hello --info=xxxx The value of the --info= option passed in the command, and shows: xxxx

For task tasks or plugins that are not a main entry, you can use this:

task("hello")
    on_run(function ())
        import("core.base.option")
        print(option.get("info"))
    end)

core.base.global

Used to get the configuration information of xmake global, that is, the value of the parameter option passed in xmake g|global --xxx=val.

!> Prior to version 2.1.5, it was core.project.global.

global.get

Similar to config.get, the only difference is that this is obtained from the global configuration.

global.load

Similar to global.get, the only difference is that this is loaded from the global configuration.

global.directory

The default is the ~/.config directory.

global.dump

The output is as follows:

{
    clean = true
,   ccache = "ccache"
,   xcode_dir = "/Applications/Xcode.app"
}

core.base.task

Used for task operations, generally used to call other task tasks in custom scripts and plug-in tasks.

!> Prior to version 2.1.5, it was core.project.task.

task.run

Used to run tasks or plugins defined by task in custom scripts, plugin tasks, for example:

task("hello")
    on_run(function ()
        print("hello xmake!")
    end)

target("demo")
    on_clean(function(target)

        -- Import task module
        import("core.base.task")

        -- Run this hello task
        task.run("hello")
    end)

We can also increase parameter passing when running a task, for example:

task("hello")
    on_run(function (arg1, arg2)
        print("hello xmake: %s %s!", arg1, arg2)
    end)

target("demo")
    on_clean(function(target)

        -- Import task
        import("core.base.task")

        -- {} This is used for the first option, which is set to null, where two arguments are passed in the last: arg1, arg2
        task.run("hello", {}, "arg1", "arg2")
    end)

The second argument to task.run is used to pass options from the command line menu instead of passing directly into the function (arg, ...) function entry, for example:

-- Import task
import("core.base.task")

-- Plugin entry
function main(...)

    -- Run the built-in xmake configuration task, equivalent to: xmake f|config --plat=iphoneos --arch=armv7
    task.run("config", {plat="iphoneos", arch="armv7"})
end

core.base.json

xmake provides a built-in json module, based on the implementation of lua_cjson, we can use it to quickly and directly interoperate between json and lua table.

We can use import("core.base.json") for direct import and use.

There are also some examples here: Jsom Examples

json.decode

import("core.base.json")
local luatable = json.decode('[1,"2", {"a":1, "b":true}]')
print(luatable)
{
    1.0,
    "2",
    {
      b = true,
      a = 1.0
    }
  }

!> If there is null in it, you can use json.null to judge

json.encode

local jsonstr = json.encode({1, "2", {a = 1}})

It should be noted that if you need to encode null, you need to use json.null, for example

local jsonstr = json.encode({json.null, 1, "2", false, true})

json.loadfile

local luatable = json.loadfile("/tmp/xxx.json")

json.savefile

json.savefile("/tmp/xxx.json", {1, {a = 1}})

core.base.semver

A module to work with semantic versionning (semver). It allows you to parse version strings and access various components of a version such as major, minor, ...

We can use import("core.base.semver") for direct import and use.

semver.new

local version = semver.new("v2.1.0")

semver.try_parse

local version = semver.try_parse("v2.1.0")

semver.match

print(semver.match("v2.1.0", 1)) -- start from position 1
print(semver.match("v2.1.0", 0, "%d+%.%d+"))
2.1.0
2.1

semver.is_valid

print(semver.is_valid("536f2bd6a092eba91315b7d1e120dff63392a11d"))
print(semver.is_valid("v2.1.0-pre"))
false
true

semver.is_valid_range

print(semver.is_valid_range(">2.1.0"))
print(semver.is_valid_range(">v2.1.0"))
print(semver.is_valid_range("2.0.0 - <2.1.0"))
print(semver.is_valid_range("1.0 || 2.1"))
true
false
false
true

semver.compare

print(semver.compare("v2.2", "v2.2.0"))
print(semver.compare("v2.2.0", "v2.1.0"))
print(semver.compare("v2.1.1", "v2.1.0"))
print(semver.compare("v2.1.0", "v2.2.0"))
0
1
1
-1

semver.satisfies

print(semver.satisfies("v2.1.0", ">= 2.1"))
print(semver.satisfies("v2.1.0", ">1.0 <2.1"))
print(semver.satisfies("v2.1.0", ">1.0 || <2.1"))
print(semver.satisfies("v2.1.0", ">=2.x || 3.x - 4.x"))
true
false
true
true

semver.select

local version, source = semver.select(">=1.5.0 <1.6", {"1.5.0", "1.5.1"}, {"v1.5.0"}, {"master", "dev"})
print(semver.select(">=1.5.0 <1.6", {"1.5.0", "1.5.1"}, {"v1.5.0"}, {"master", "dev"}))
print(semver.select("v1.5.0", {"1.5.0", "1.5.1"}, {"v1.5.0"}, {"master", "dev"}))
print(semver.select("master", {"1.5.0", "1.5.1"}, {"v1.5.0"}, {"master", "dev"}))
1.5.1 version
v1.5.0 tag
master branch

semver:get

local version = semver.new("v2.1.0+build")
print(version["_INFO"])
print(version:get("major"))
{ 
  prerelease = { },
  build = {
    "build"
  },
  version = "v2.1.0+build",
  raw = "v2.1.0+build",
  patch = 0,
  minor = 1,
  major = 2
}

2

semver:major

semver.new("v2.1.0"):major() -- return 2

semver:minor

semver.new("v2.1.0"):minor() -- return 1

semver:patch

semver.new("v2.1.0"):patch() -- return 0

semver:build

semver.new("v2.1.0+build"):build() -- return {"build"}

semver:prerelease

semver.new("v2.1.0-prerelease"):prerelease() -- return {"prerelease"}

semver:rawstr

semver.new("v2.1.0+build"):rawstr() -- return v2.1.0+build

semver:shortstr

semver.new("v2.1.0+build"):shortstr() -- return 2.1.0

core.tool.linker

Linker related operations, often used for plugin development.

For the target, link the specified object file list to generate the corresponding target file, for example:

linker.link("binary", "cc", {"a.o", "b.o", "c.o"}, target:targetfile(), {target = target})

Where target is the project target, here is passed in, mainly used to get the target-specific link options. For the project target object, see: core.project.project

Of course, you can also not specify the target, for example:

linker.link("binary", "cc", {"a.o", "b.o", "c.o"}, "/tmp/targetfile")

The first parameter specifies the link type and currently supports: binary, static, shared
The second parameter tells the linker that it should be linked as the source file object, and what compiler source files are compiled with, for example:

Second Parameter Value Description
cc c compiler
cxx c++ compiler
mm objc compiler
mxx objc++ compiler
gc go compiler
as assembler
sc swift compiler
rc rust compiler
dc dlang compiler

Specifying different compiler types, the linker will adapt the most appropriate linker to handle the link, and if several languages ​​support mixed compilation, you can pass in multiple compiler types at the same time, specifying that the linker chooses to support these hybrid compilations. The linker of the language performs link processing:

linker.link("binary", {"cc", "mxx", "sc"}, {"a.o", "b.o", "c.o"}, "/tmp/targetfile")

The above code tells the linker that the three object files a, b, c may be c, objc++, compiled by swift code. The linker will select the most suitable linker from the current system and toolchain to handle the link process. .

linker.linkcmd

Get the command line string executed in linker.link directly, which is equivalent to:

local cmdstr = linker.linkcmd("static", "cxx", {"a.o", "b.o", "c.o"}, target:targetfile(), {target = target})

Note: The extension part of target = target}` is optional. If the target object is passed, the generated link command will add the link option corresponding to this target configuration.

And you can also pass various configurations yourself, for example:

local cmdstr = linker.linkcmd("static", "cxx", {"a.o", "b.o", "c.o"}, target:targetfile(), {configs = {linkdirs = "/usr/lib"}})

linker.linkargv

A little different from linker.linkcmd is that this interface returns a list of parameters, table representation, more convenient to operate:

local program, argv = linker.linkargv("static", "cxx", {"a.o", "b.o", "c.o"}, target:targetfile(), {target = target})

The first value returned is the main program name, followed by the parameter list, and os.args(table.join(program, argv)) is equivalent to linker.linkcmd.

We can also run it directly by passing the return value to os.runv: os.runv(linker.linkargv(..))

linker.linkflags

Get the link option string part of linker.linkcmd without shellname and object file list, and return by array, for example:

local flags = linker.linkflags("shared", "cc", {target = target})
for _, flag in ipairs(flags) do
    print(flag)
end

The returned array of flags is an array.

linker.has_flags

Although it can be judged by lib.detect.has_flags, but the interface is more low-level, you need to specify the linker name.
This interface only needs to specify the target type of the target, the source file type, which will automatically switch to select the currently supported linker.

if linker.has_flags(target:targetkind(), target:sourcekinds(), "-L/usr/lib -lpthread") then
    -- ok
end

core.tool.compiler

Compiler related operations, often used for plugin development.

compiler.compile

For the target, link the specified object file list to generate the corresponding target file, for example:

compiler.compile("xxx.c", "xxx.o", "xxx.h.d", {target = target})

Where target is the project target, here is the specific compile option that is mainly used to get the target. If you get the project target object, see: core.project.project

The xxx.h.d file is used to store the header file dependency file list for this source file. Finally, these two parameters are optional. You can not pass them when compiling:

compiler.compile("xxx.c", "xxx.o")

To simply compile a source file.

compiler.compcmd

Get the command line string executed directly in compiler.compile, which is equivalent to:

local cmdstr = compiler.compcmd("xxx.c", "xxx.o", {target = target})

Note: The extension part of target = target}` is optional. If the target object is passed, the generated compile command will add the link option corresponding to this target configuration.

And you can also pass various configurations yourself, for example:

local cmdstr = compiler.compcmd("xxx.c", "xxx.o", {configs = {includedirs = "/usr/include", defines = "DEBUG"}})

With target, we can export all source file compilation commands for the specified target:

import("core.project.project")

for _, target in pairs(project.targets()) do
    for sourcekind, sourcebatch in pairs(target:sourcebatches()) do
        for index, objectfile in ipairs(sourcebatch.objectfiles) do
            local cmdstr = compiler.compcmd(sourcebatch.sourcefiles[index], objectfile, {target = target})
        end
    end
end

compiler.compargv

A little different from compiler.compcmd is that this interface returns a list of parameters, table representation, more convenient to operate:

local program, argv = compiler.compargv("xxx.c", "xxx.o")

compiler.compflags

Get the compile option string part of compiler.compcmd without shList of ellnames and files, for example:

local flags = compiler.compflags(sourcefile, {target = target})
for _, flag in ipairs(flags) do
    print(flag)
end

The returned array of flags is an array.

compiler.has_flags

Although it can be judged by lib.detect.has_flags, but the interface is more low-level, you need to specify the compiler name.
This interface only needs to specify the language type, it will automatically switch to select the currently supported compiler.

-- Determine if the c language compiler supports the option: -g
if compiler.has_flags("c", "-g") then
    -- ok
end

-- Determine if the C++ language compiler supports the option: -g
if compiler.has_flags("cxx", "-g") then
    -- ok
end

compiler.features

Although it can be obtained by lib.detect.features, but the interface is more low-level, you need to specify the compiler name.
This interface only needs to specify the language type, it will automatically switch to select the currently supported compiler, and then get the current list of compiler features.

-- Get all the features of the current c compiler
local features = compiler.features("c")

-- Get all the features of the current C++ language compiler, enable the C++11 standard, otherwise you will not get the new standard features.
local features = compiler.features("cxx", {cofnig = {cxxflags = "-std=c++11"}})

-- Get all the features of the current C++ language compiler, pass all configuration information of the project target
local features = compiler.features("cxx", {target = target, configs = {defines = "..", includedirs = ".."}})

A list of all c compiler features:

Feature Name
c_static_assert
c_restrict
c_variadic_macros
c_function_prototypes

A list of all C++ compiler features:

Feature Name
cxx_variable_templates
cxx_relaxed_constexpr
cxx_aggregate_default_initializers
cxx_contextual_conversions
cxx_attribute_deprecated
cxx_decltype_auto
cxx_digit_separators
cxx_generic_lambdas
cxx_lambda_init_captures
cxx_binary_literals
cxx_return_type_deduction
cxx_decltype_incomplete_return_types
cxx_reference_qualified_functions
cxx_alignof
cxx_attributes
cxx_inheriting_constructors
cxx_thread_local
cxx_alias_templates
cxx_delegating_constructors
cxx_extended_friend_declarations
cxx_final
cxx_nonstatic_member_init
cxx_override
cxx_user_literals
cxx_constexpr
cxx_defaulted_move_initializers
cxx_enum_forward_declarations
cxx_noexcept
cxx_nullptr
cxx_range_for
cxx_unrestricted_unions
cxx_explicit_conversions
cxx_lambdas
cxx_local_type_template_args
cxx_raw_string_literals
cxx_auto_type
cxx_defaulted_functions
cxx_deleted_functions
cxx_generalized_initializers
cxx_inline_namespaces
cxx_sizeof_member
cxx_strong_enums
cxx_trailing_return_types
cxx_unicode_literals
cxx_uniform_initialization
cxx_variadic_templates
cxx_decltype
cxx_default_function_template_args
cxx_long_long_type
cxx_right_angle_brackets
cxx_rvalue_references
cxx_static_assert
cxx_extern_templates
cxx_func_identifier
cxx_variadic_macros
cxx_template_template_parameters

compiler.has_features

Although it can be obtained by lib.detect.has_features, but the interface is more low-level, you need to specify the compiler name.
And this interface only needs to specify the special name list that needs to be detected, it can automatically switch to select the currently supported compiler, and then determine whether the specified feature is supported in the current compiler.

if compiler.has_features("c_static_assert") then
    -- ok
end

if compiler.has_features({"c_static_assert", "cxx_constexpr"}, {languages ​​= "cxx11"}) then
    -- ok
end

if compiler.has_features("cxx_constexpr", {target = target, defines = "..", includedirs = ".."}) then
    -- ok
end

For specific feature names, refer to compiler.features.

core.project.config

Used to get the configuration information when the project is compiled, that is, the value of the parameter option passed in xmake f|config --xxx=val.

config.get

Used to get the configuration value of xmake f|config --xxx=val, for example:

target("test")
    on_run(function (target)

        -- Import configuration module
        import("core.project.config")

        -- Get configuration values
        print(config.get("xxx"))
    end)

config.load

Generally used in plug-in development, the plug-in task is not like the custom script of the project, the environment needs to be initialized and loaded by itself, the default project configuration is not loaded, if you want to use config.get interface to get the project Configuration, then you need to:


-- Import configuration module
import("core.project.config")

function main(...)

    -- Load project configuration first
    config.load()

    -- Get configuration values
    print(config.get("xxx"))
end

config.arch

That is to get the platform configuration of xmake f|config --arch=armv7, which is equivalent to config.get("arch").

config.plat

That is to get the platform configuration of xmake f|config --plat=iphoneos, which is equivalent to config.get("plat").

config.mode

That is to get the platform configuration of xmake f|config --mode=debug, which is equivalent to config.get("mode").

config.buildir

That is to get the platform configuration of xmake f|config -o /tmp/output, which is equivalent to config.get("buildir").

config.directory

Get the storage directory of the project configuration, the default is: projectdir/.config

config.dump

The output is for example:

{
    sh = "xcrun -sdk macosx clang++"
,   xcode_dir = "/Applications/Xcode.app"
,   ar = "xcrun -sdk macosx ar"
,   small = true
,   object = false
,   arch = "x86_64"
,   xcode_sdkver = "10.12"
,   ex = "xcrun -sdk macosx ar"
,   cc = "xcrun -sdk macosx clang"
,   rc = "rustc"
,   plat = "macosx"
,   micro = false
,   host = "macosx"
,   as = "xcrun -sdk macosx clang"
,   dc = "dmd"
,   gc = "go"
,   openssl = false
,   ccache = "ccache"
,   cxx = "xcrun -sdk macosx clang"
,   sc = "xcrun -sdk macosx swiftc"
,   mm = "xcrun -sdk macosx clang"
,   buildir = "build"
,   mxx = "xcrun -sdk macosx clang++"
,   ld = "xcrun -sdk macosx clang++"
,   mode = "release"
,   kind = "static"
}

core.project.project

Used to get some description information of the current project, that is, the configuration information defined in the xmake.lua project description file, for example: target, option, etc.

project.load

It is only used in the plugin, because the project configuration information has not been loaded at this time. In the custom script of the project target, you do not need to perform this operation, you can directly access the project configuration.

-- Import engineering modules
import("core.project.project")

-- Plugin entry
function main(...)

    -- Load project description configuration
    project.load()

    -- access project descriptions, such as getting specified project goals
    local target = project.target("test")
end

!> After version 2.1.5, if not needed, the project load will automatically load at the appropriate time.

project.directory

Get the current project directory, which is the directory specified in xmake -P xxx, otherwise it is the default current xmake command execution directory.

!> After version 2.1.5, it is recommended to use os.projectdir to get it.

project.target

Get and access the specified project target configuration, for example:

local target = project.target("test")
if target then

    -- Get the target file name
    print(target:targetfile())

    -- Get the target type, which is: binary, static, shared
    print(target:targetkind())

    -- Get the target name
    print(target:name())

    -- Get the target source file
    local sourcefiles = target:sourcefiles()

    -- Get a list of target installation header files
    local srcheaders, dstheaders = target:headerfiles()

    -- Get target dependencies
    print(target:get("deps"))
end

project.targets

Returns all compilation targets for the current project, for example:

for targetname, target in pairs(project.targets()) do
    print(target:targetfile())
end

project.option

Get and access the option objects specified in the project, for example:

local option = project.option("test")
if option:enabled() then
    option:enable(false)
end

project.options

Returns all compilation targets for the current project, for example:

for optionname, option in pairs(project.options())
    print(option:enabled())
end

project.name

That is, get the project name configuration of set_project.

print(project.name())

project.version

That is, get set_version project version configuration.

print(project.version())

core.language.language

Used to obtain information about the compiled language, generally used for the operation of code files.

language.extensions

The results are as follows:

{
     [".c"] = cc
,    [".cc"] = cxx
,    [".cpp"] = cxx
,    [".m"] = mm
,    [".mm"] = mxx
,    [".swift"] = sc
,    [".go"] = gc
}

language.targetkinds

The results are as follows:

{
     binary = {"ld", "gcld", "dcld"}
,    static = {"ar", "gcar", "dcar"}
,    shared = {"sh", "dcsh"}
}

language.sourcekinds

The results are as follows:

{
     cc = ".c"
,    cxx = {".cc", ".cpp", ".cxx"}
,    mm = ".m"
,    mxx = ".mm"
,    sc = ".swift"
,    gc = ".go"
,    rc = ".rs"
,    dc = ".d"
,    as = {".s", ".S", ".asm"}
}

language.sourceflags

The results are as follows:

{
     cc = {"cflags", "cxflags"}
,    cxx = {"cxxflags", "cxflags"}
,    ...
}

language.load

Load a specific language object from the language name, for example:

local lang = language.load("c++")
if lang then
    print(lang:name())
end

language.load_sk

Load specific language objects from the source file type: cc, cxx, mm, mxx, sc, gc, as .., for example:

local lang = language.load_sk("cxx")
if lang then
    print(lang:name())
end

language.load_ex

Load specific language objects from the source file extension: .cc, .c, .cpp, .mm, .swift, .go .., for example:

local lang = language.load_ex(".cpp")
if lang then
    print(lang:name())
end

language.sourcekind_of

That is, from a given source file path, get the type of source file it belongs to, for example:

print(language.sourcekind_of("/xxxx/test.cpp"))

The result is: cxx, which is the c++ type. For the corresponding list, see: language.sourcekinds

lib.detect

This module provides very powerful probing capabilities for probing programs, compilers, language features, dependencies, and more.

!> The interface of this module is spread across multiple module directories, try to import it by importing a single interface, which is more efficient.

detect.find_file

This interface provides a more powerful project than os.files, which can specify multiple search directories at the same time, and can also specify additional subdirectories for each directory to match the pattern lookup, which is equivalent to An enhanced version of os.files.

E.g:

import("lib.detect.find_file")

local file = find_file("ccache", { "/usr/bin", "/usr/local/bin"})

If found, the result returned is: /usr/bin/ccache

It also supports pattern matching paths for recursive lookups, similar to os.files:

local file = find_file("test.h", { "/usr/include", "/usr/local/include/**"})

Not only that, but the path inside also supports built-in variables to get the path from the environment variables and the registry to find:

local file = find_file("xxx.h", { "$(env PATH)", "$(reg HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\XXXX;Name)"})

If the path rules are more complex, you can also dynamically generate path entries through a custom script:

local file = find_file("xxx.h", { "$(env PATH)", function () return val("HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\XXXX;Name"):match ("\"(.-)\"") end})

In most cases, the above use has met various needs. If you need some extended functions, you can customize some optional configurations by passing in the third parameter, for example:

local file = find_file("test.h", { "/usr", "/usr/local"}, {suffixes = {"/include", "/lib"}})

By specifying a list of suffixes subdirectories, you can extend the list of paths (the second parameter) so that the actual search directory is expanded to:

/usr/include
/usr/lib
/usr/local/include
/usr/local/lib

And without changing the path list, you can dynamically switch subdirectories to search for files.

!> We can also quickly call and test this interface with the xmake lua plugin: xmake lua lib.detect.find_file test.h /usr/local

detect.find_path

The usage of this interface is similar to lib.detect.find_file, the only difference is that the returned results are different.
After the interface finds the incoming file path, it returns the corresponding search path, not the file path itself. It is generally used to find the parent directory location corresponding to the file.

import("lib.detect.find_path")

local p = find_path("include/test.h", { "/usr", "/usr/local"})

If the above code is successful, it returns: /usr/local, if test.h is in /usr/local/include/test.h.

Another difference is that this interface is passed in not only the file path, but also the directory path to find:

local p = find_path("lib/xxx", { "$(env PATH)", "$(reg HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\XXXX;Name)"})

Again, this interface also supports pattern matching and suffix subdirectories:

local p = find_path("include/*.h", { "/usr", "/usr/local/**"}, {suffixes = "/subdir"})

detect.find_library

This interface is used to find library files (static libraries, dynamic libraries) in the specified search directory, for example:

import("lib.detect.find_library")

local library = find_library("crypto", {"/usr/lib", "/usr/local/lib"})

Running on macosx, the results returned are as follows:

{
    filename = libcrypto.dylib
,   linkdir = /usr/lib
,   link = crypto
,   kind = shared
}

If you do not specify whether you need a static library or a dynamic library, then this interface will automatically select an existing library (either a static library or a dynamic library) to return.

If you need to force the library type you need to find, you can specify the kind parameter as (static/shared):

local library = find_library("crypto", {"/usr/lib", "/usr/local/lib"}, {kind = "static"})

This interface also supports suffixes suffix subdirectory search and pattern matching operations:

local library = find_library("cryp*", {"/usr", "/usr/local"}, {suffixes = "/lib"})

detect.find_program

This interface is more primitive than lib.detect.find_tool, looking for executables through the specified parameter directory.

import("lib.detect.find_program")

local program = find_program("ccache")

The above code is like not passing the search directory, so it will try to execute the specified program directly. If it runs ok, it will return directly: ccache, indicating that the search is successful.

Specify the search directory and modify the test command parameters that are attempted to run (default: ccache --version):

localProgram = find_program("ccache", {paths = {"/usr/bin", "/usr/local/bin"}, check = "--help"})

The above code will try to run: /usr/bin/ccache --help, if it runs successfully, it returns: /usr/bin/ccache.

If --help can't satisfy the requirement, some programs don't have the --version/--help parameter, then you can customize the run script to run the test:

local program = find_program("ccache", {paths = {"/usr/bin", "/usr/local/bin"}, check = function (program) os.run("%s -h", program) end })

Similarly, the search path list supports built-in variables and custom scripts:

local program = find_program("ccache", {paths = {"$(env PATH)", "$(reg HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\AeDebug;Debugger)"}})
local program = find_program("ccache", {paths = {"$(env PATH)", function () return "/usr/local/bin" end}})

!> In order to speed up the efficiency of frequent lookups, this interface comes with a default cache, so even if you frequently find the same program, it will not take too much time.
If you want to disable the cache, you can clear the local cache by executing xmake f -c in the project directory.

We can also test quickly with xmake lua lib.detect.find_program ccache.

detect.find_programver

import("lib.detect.find_programver")

local programver = find_programver("ccache")

The return result is: 3.2.2

By default it will try to get the version via ccache --version. If this parameter doesn't exist, you can specify other parameters yourself:

local version = find_programver("ccache", {command = "-v"})

Even the custom version gets the script:

local version = find_programver("ccache", {command = function () return os.iorun("ccache --version") end})

For the extraction rule of the version number, if the built-in matching mode does not meet the requirements, you can also customize:

local version = find_programver("ccache", {command = "--version", parse = "(%d+%.?%d*%.?%d*.-)%s"})
local version = find_programver("ccache", {command = "--version", parse = function (output) return output:match("(%d+%.?%d*%.?%d*.-)%s ") end})

!> In order to speed up the efficiency of frequent lookups, this interface is self-contained by default. If you want to disable the cache, you can execute xmake f -c in the project directory to clear the local cache.

We can also test quickly with xmake lua lib.detect.find_programver ccache.

detect.find_package

After 2.6.x this interface is not recommended for direct use (internal use only), for library integration, please use add_requires() and add_packages() as much as possible.

detect.find_tool

This interface is also used to find executable programs, but more
advanced than lib.detect.find_program, the
function is also more powerful, it encapsulates the executable
program, providing the concept of tools:

The corresponding relationship is as follows:

toolname program
clang xcrun -sdk macosx clang
gcc /usr/toolchains/bin/arm-linux-gcc
link link.exe -lib

lib.detect.find_program can only determine
whether the program exists by passing in the original program command
or path. And find_tool can find the tool through a more consistent
toolname, and return the corresponding program complete command path,
for example:

import("lib.detect.find_tool")

local tool = find_tool("clang")

The result returned is: {name = "clang", program = "clang"}, at this
time there is no difference, we can manually specify the executable
command:

local tool = find_tool("clang", {program = "xcrun -sdk macosx clang"})

The result returned is: {name = "clang", program = "xcrun -sdk macosx clang"}

In macosx, gcc is clang. If we execute gcc --version, we can see
that it is a vest of clang. We can intelligently identify it through
the find_tool interface:

local tool = find_tool("gcc")

The result returned is: {name = "clang", program = "gcc"}

The difference can be seen by this result. The tool name will actually
be marked as clang, but the executable command uses gcc.

We can also specify the {version = true} parameter to get the
version of the tool, and specify a custom search path. It also
supports built-in variables and custom scripts:

local tool = find_tool("clang", {version = true, paths = {"/usr/bin", "/usr/local/bin", "$(env PATH)", function () return "/usr/xxx/bin" end}})

The result returned is: {name = "clang", program = "/usr/bin/clang", version = "4.0"}

This interface is a high-level wrapper around find_program, so it
also supports custom script detection:

local tool = find_tool("clang", {check = "--help"})
local tool = find_tool("clang", {check = function (tool) os.run("%s -h", tool) end})

Finally, the search process of find_tool:

  1. First try to run and detect with the argument of {program = "xxx"}. 2. If there is a detect.tools.find_xxx script in
    xmake/modules/detect/tools, call this script for more accurate
    detection. 3. Try to detect from the system directory such as
    /usr/bin, /usr/local/bin.

We can also add a custom lookup script to the module directory
specified by add_moduledirs in the project xmake.lua to improve
the detection mechanism:

projectdir
- xmake/modules
- detect/tools/find_xxx.lua

For example, we customize a lookup script for find_7z.lua:

import("lib.detect.find_program")
import("lib.detect.find_programver")

function main(opt)

    -- init options
    opt = opt or {}

    -- find program
    local program = find_program(opt.program or "7z", opt.pathes, opt.check or "--help")

    -- find program version
    local version = nil
    if program and opt and opt.version then
        version = find_programver(program, "--help", "(%d+%.?%d*)%s")
    end

    -- ok?
    return program, version
end

After placing it in the project's module directory, execute: xmake l lib.detect.find_tool 7z to find it.

!> In order to speed up the efficiency of frequent
lookups, this interface is self-contained by default. If you want to
disable the cache, you can execute xmake f -c in the project
directory to clear the local cache.

We can also test quickly with xmake lua lib.detect.find_tool clang.

detect.find_toolname

Match the corresponding tool name with the program command, for
example:

program toolname
xcrun -sdk macosx clang clang
/usr/bin/arm-linux-gcc gcc
link.exe -lib link
gcc-5 gcc
arm-android-clang++ clangxx
pkg-config pkg_config

Compared with program, toolname can uniquely mark a tool, and it is also convenient to find and load the corresponding script find_xxx.lua.

detect.find_cudadevices

Enumerate CUDA devices through the CUDA Runtime API and query theirs properties.

import("lib.detect.find_cudadevices")

local devices = find_cudadevices({ skip_compute_mode_prohibited = true })
local devices = find_cudadevices({ min_sm_arch = 35, order_by_flops = true })

The result returned is: { { ['$id'] = 0, name = "GeForce GTX 960M", major = 5, minor = 0, ... }, ... }

The included properties will vary depending on the current CUDA version.
Please refer to CUDA Toolkit Documentation and its historical version for more information.

detect.features

This interface is similar to compiler.features. The difference is that this interface is more primitive. The passed argument is the actual tool name toolname.

And this interface not only can get the characteristics of the compiler, the characteristics of any tool can be obtained, so it is more versatile.

import("lib.detect.features")

local features = features("clang")
local features = features("clang", {flags = "-O0", program = "xcrun -sdk macosx clang"})
local features = features("clang", {flags = {"-g", "-O0", "-std=c++11"}})

By passing in flags, you can change the result of the feature, for example, some features of C++11, which are not available by default. After enabling -std=c++11, you can get it.

A list of all compiler features can be found at compiler.features.

detect.has_features

This interface is similar to compiler.has_features, but more primitive, the passed argument is the actual tool name toolname.

And this interface can not only judge the characteristics of the compiler, but the characteristics of any tool can be judged, so it is more versatile.

import("lib.detect.has_features")

local features = has_features("clang", "cxx_constexpr")
local features = has_features("clang", {"cxx_constexpr", "c_static_assert"}, {flags = {"-g", "-O0"}, program = "xcrun -sdk macosx clang"})
local features = has_features("clang", {"cxx_constexpr", "c_static_assert"}, {flags = "-g"})

If the specified feature list exists, the actual supported feature sublist is returned. If none is supported, nil is returned. We can also change the feature acquisition rule by specifying flags.

A list of all compiler features can be found at compiler.features.

detect.has_flags

This interface is similar to compiler.has_flags, but more primitive, the passed argument is the actual tool name toolname.

import("lib.detect.has_flags")

local ok = has_flags("clang", "-g")
local ok = has_flags("clang", {"-g", "-O0"}, {program = "xcrun -sdk macosx clang"})
local ok = has_flags("clang", "-g -O0", {toolkind = "cxx"})

Returns true if the test passed.

The detection of this interface has been optimized. Except for the cache mechanism, in most cases, the tool's option list (--help) will be directly judged. If the option list is not available, it will be tried. The way to run to detect.

detect.has_cfuncs

This interface is a simplified version of lib.detect.check_cxsnippets and is only used to detect functions.

import("lib.detect.has_cfuncs")

local ok = has_cfuncs("setjmp")
local ok = has_cfuncs({"sigsetjmp((void*)0, 0)", "setjmp"}, {includes = "setjmp.h"})

The rules for describing functions are as follows:

Function Description Description
sigsetjmp pure function name
sigsetjmp((void*)0, 0) Function Call
sigsetjmp{int a = 0; sigsetjmp((void*)a, a);} function name + {} block

In the last optional parameter, in addition to specifying includes, you can also specify other parameters to control the option conditions for compile detection:

{ verbose = false, target = [target|option], includes = .., configs = {linkdirs = .., links = .., defines = ..}}

The verbose is used to echo the detection information, the target is used to append the configuration information in the target before the detection, and the config is used to customize the compilation options related to the target.

detect.has_cxxfuncs

This interface is similar to lib.detect.has_cfuncs, please refer to its instructions for use. The only difference is that this interface is used to detect c++ functions.

detect.has_cincludes

This interface is a simplified version of lib.detect.check_cxsnippets and is only used to detect header files.

import("lib.detect.has_cincludes")

local ok = has_cincludes("stdio.h")
local ok = has_cincludes({"stdio.h", "stdlib.h"}, {target = target})
local ok = has_cincludes({"stdio.h", "stdlib.h"}, {configs = {defines = "_GNU_SOURCE=1", languages ​​= "cxx11"}})

detect.has_cxxincludes

This interface is similar to lib.detect.has_cincludess, please refer to its instructions for use. The only difference is that this interface is used to detect c++ header files.

detect.has_ctypes

This interface is a simplified version of lib.detect.check_cxsnippets and is only used to detect functions.

import("lib.detect.has_ctypes")

local ok = has_ctypes("wchar_t")
local ok = has_ctypes({"char", "wchar_t"}, {includes = "stdio.h"})
local ok = has_ctypes("wchar_t", {includes = {"stdio.h", "stdlib.h"}, configs = {"defines = "_GNU_SOURCE=1", languages ​​= "cxx11"}})

detect.has_cxxtypes

This interface is similar to lib.detect.has_ctypess. Please refer to its instructions for use. The only difference is that this interface is used to detect c++ types.

detect.check_cxsnippets

The generic c/c++ code snippet detection interface, by passing in a list of multiple code snippets, it will automatically generate a compiled file, and then common sense to compile it, if the compilation pass returns true.

For some complex compiler features, even if compiler.has_features can't detect it, you can detect it by trying to compile through this interface.

import("lib.detect.check_cxsnippets")

local ok = check_cxsnippets("void test() {}")
local ok = check_cxsnippets({"void test(){}", "#define TEST 1"}, {types = "wchar_t", includes = "stdio.h"})

This interface is a generic version of interfaces such as detect.has_cfuncs, detect.has_cincludes, and detect.has_ctypes, and is also lower level.

So we can use it to detect: types, functions, includes and links, or combine them together to detect.

The first parameter is a list of code fragments, which are generally used for the detection of some custom features. If it is empty, it can only detect the conditions in the optional parameters, for example:

local ok = check_cxsnippets({}, {types = {"wchar_t", "char*"}, includes = "stdio.h", funcs = {"sigsetjmp", "sigsetjmp((void*)0, 0)"} })

The above call will check if the types, includes and funcs are both satisfied, and return true if passed.

There are other optional parameters:

{ verbose = false, target = [target|option], sourcekind = "[cc|cxx]"}

The verbose is used to echo the detection information. The target is used to append the configuration information in the target before the detection. The sourcekind is used to specify the tool type such as the compiler. For example, the incoming cxx is forced to be detected as c++ code.

net.http

This module provides various operational support for http. The currently available interfaces are as follows:

http.download

This interface is relatively simple, is simply download files.

import("net.http")

http.download("https://xmake.io", "/tmp/index.html")

privilege.sudo

This interface is used to run commands via sudo and provides platform consistency handling, which can be used for scripts that require root privileges to run.

!> In order to ensure security, unless you must use it, try not to use this interface in other cases.

sudo.has

At present, sudo is supported only under macosx/linux. The administrator privilege running on Windows is not supported yet. Therefore, it is recommended to use the interface to judge the support situation before use.

import("privilege.sudo")

if sudo.has() then
    sudo.run("rm /system/file")
end

sudo.run

For specific usage, please refer to: os.run.

import("privilege.sudo")

sudo.run("rm /system/file")

sudo.runv

For specific usage, please refer to: os.runv.

sudo.exec

For specific usage, please refer to: os.exec.

sudo.execv

For specific usage, please refer to: os.execv.

sudo.iorun

For specific usage, please refer to: os.iorun.

sudo.iorunv

For specific usage, please refer to: os.iorunv.

devel.git

This interface provides access to various commands of git. Compared to the direct call to git command, this module provides a more easy-to-use package interface and provides automatic detection and cross-platform processing for git.

!> Currently on Windows, you need to manually install the git package before you can detect it. The subsequent version will provide automatic integration of git function. Users will not need to care about how to install git, they can be used directly.

git.clone

This interface corresponds to the git clone command.

import("devel.git")

git.clone("git@github.com:tboox/xmake.git")
git.clone("git@github.com:tboox/xmake.git", {depth = 1, branch = "master", outputdir = "/tmp/xmake"})

git.pull

This interface corresponds to the git pull command.

import("devel.git")

git.pull()
git.pull({remote = "origin", tags = true, branch = "master", repodir = "/tmp/xmake"})

git.clean

This interface corresponds to the git clean command.

import("devel.git")

git.clean()
git.clean({repodir = "/tmp/xmake", force = true})

git.checkout

This interface corresponds to the git checkout command

import("devel.git")

git.checkout("master", {repodir = "/tmp/xmake"})
git.checkout("v1.0.1", {repodir = "/tmp/xmake"})

git.refs

This interface corresponds to the git ls-remote --refs command

import("devel.git")

local refs = git.refs(url)

git.tags

This interface corresponds to the git ls-remote --tags command

import("devel.git")

local tags = git.tags(url)

git.branches

This interface corresponds to the git ls-remote --heads command

import("devel.git")

local branches = git.branches(url)

utils.archive

This module is used to compress and decompress files. It supports decompression of most common compression formats. It will automatically detect which compression tools are provided by the system, and then will use the most appropriate compression tool for the operation.

archive.archive

import("utils.archive")

archive.archive("/tmp/a.zip", "/tmp/outputdir")
archive.archive("/tmp/a.7z", "/tmp/outputdir")
archive.archive("/tmp/a.gzip", "/tmp/outputdir")
archive.archive("/tmp/a.tar.bz2", "/tmp/outputdir")

Some configuration options can also be added, such as recursive directories, compression quality, exclude files, etc.

import("utils.archive")

local options = {}
options.curdir = "/tmp"
options.recurse = true
options.compress = "fastest|faster|default|better|best"
options.excludes = {"*/dir/*", "dir/*"}
archive.archive("/tmp/a.zip", "/tmp/outputdir", options)

archive.extract

import("utils.archive")

archive.extract("/tmp/a.zip", "/tmp/outputdir")
archive.extract("/tmp/a.7z", "/tmp/outputdir")
archive.extract("/tmp/a.gzip", "/tmp/outputdir")
archive.extract("/tmp/a.tar.bz2", "/tmp/outputdir")

utils.platform

This module is used for some platform-related auxiliary operation interfaces

utils.platform.gnu2mslib

This feature is particularly helpful for Fortran & C++ mixed projects, because VS does not provide the fortran compiler. You can only use MinGW's gfortran to compile the fortran part, and then link with the VS project.
Often such projects also have some other libraries provided in the vs format, so pure MinGW compilation is not possible, you can only use this function to mix compilation.

And cmake also has a similar GNUtoMS.

For related issues, see: #1181

import("utils.platform.gnu2mslib")

gnu2mslib("xxx.lib", "xxx.dll.a")
gnu2mslib("xxx.lib", "xxx.def")
gnu2mslib("xxx.lib", "xxx.dll.a", {dllname = "xxx.dll", arch = "x64"})

Support to generate xxx.lib from def, and also support to automatically export .def from xxx.dll.a, and then generate xxx.lib

If you don’t want to automatically generate def from dll.a, but want to borrow the def generated by gnu linker, then configure it yourself through add_shflags("-Wl,--output-def,xxx.def") to generate def, and then pass in def to this interface. .

{dllname = xxx, arch = "xxx"} These are optional, according to your needs.

You can also directly xmake l utils.platform.gnu2mslib xxx.lib xxx.dll.a quick test verification

cli

cli.amalgamate

This is a small utility module for quickly merging all c/c++ and header sources inside a given target into a single source file.

The merge expands all the internal includes headers and generates a DAG, which is introduced by topological sorting.

By default it will handle the merge of all targets, for example:

$ xmake l cli.amalgamate
build/tbox.c generated!
build/tbox.h generated!

We can also specify the targets we need for the merge:

$ xmake l cli.amalgamate tbox
build/tbox.c generated!
build/tbox.h generated!

You can also handle symbol conflicts by specifying a custom unique ID macro definition for each source file you merge.

$ xmake l cli.amalgamate -u MY_UNIQUEU_ID
build/tbox.c generated!
build/tbox.h generated!

If there are renamed symbols in multiple source files, you can determine if the MY_UNIQUEU_ID macro is defined, and if it is, then it is in a single file, so you can handle the renamed symbols yourself in the source code.

#ifdef MY_UNIQUEU_ID
    // do some thing
#endif

We can also specify the output location:

$ xmake l cli.amalgamate -o /xxx
/xxx/tbox.c generated!
/xxx/tbox.h generated!