Attributes
name
This is a string, with a minimum of 2 and a maximum of 50 characters (though shorter names are recommended), that defines the package name. It will be the <PkgName>/version@user/channel
of the package reference.
It should match the following regex ^[a-zA-Z0-9_][a-zA-Z0-9_\+\.-]{1,50}$
, so start with alphanumeric or underscore, then alphanumeric, underscore, +, ., - characters.
The name is only necessary for export
-ing the recipe into the local cache (export
and create
commands), if they are not defined in the command line.
It might take its value from an environment variable, or even any python code that defines it (e.g. a function that reads an environment variable, or a file from disk).
However, the most common and suggested approach would be to define it in plain text as a constant, or provide it as command line arguments.
version
The version attribute will define the version part of the package reference: PkgName/<version>@user/channel
It is a string, and can take any value, matching the same constraints as the name
attribute.
In case the version follows semantic versioning in the form X.Y.Z-pre1+build2
, that value might be used for requiring this package through version ranges instead of exact versions.
The version is only strictly necessary for export
-ing the recipe into the local cache (export
and create
commands), if they are not defined in the command line.
It might take its value from an environment variable, or even any python code that defines it (e.g. a function that reads an environment variable, or a file from disk).
Please note that this value might be used in the recipe in other places (as in source()
method to retrieve code from elsewhere), making this value not constant means that it may evaluate differently in different contexts (e.g., on different machines or for different users) leading to unrepeatable or unpredictable results.
The most common and suggested approach would be to define it in plain text as a constant, or provide it as command line arguments.
description
This is an optional, but strongly recommended text field, containing the description of the package, and any information that might be useful for the consumers. The first line might be used as a short description of the package.
class HelloConan(ConanFile):
name = "Hello"
version = "0.1"
description = """This is a Hello World library.
A fully featured, portable, C++ library to say Hello World in the stdout,
with incredible iostreams performance"""
homepage
Use this attribute to indicate the home web page of the library being packaged. This is useful to link the recipe to further explanations of the library itself like an overview of its features, documentation, FAQ as well as other related information.
class EigenConan(ConanFile):
name = "eigen"
version = "3.3.4"
homepage = "http://eigen.tuxfamily.org"
url
It is possible, even typical, if you are packaging a third party lib, that you just develop
the packaging code. Such code is also subject to change, often via collaboration, so it should be stored
in a VCS like git, and probably put on GitHub or a similar service. If you do indeed maintain such a
repository, please indicate it in the url
attribute, so that it can be easily found.
class HelloConan(ConanFile):
name = "Hello"
version = "0.1"
url = "https://github.com/conan-io/hello.git"
The url
is the url of the package repository, i.e. not necessarily the original source code.
It is optional, but highly recommended, that it points to GitHub, Bitbucket or your preferred
code collaboration platform. Of course, if you have the conanfile inside your library source,
you can point to it, and afterwards use the url
in your source()
method.
This is a recommended, but not mandatory attribute.
license
This field is intended for the license of the target source code and binaries, i.e. the code
that is being packaged, not the conanfile.py
itself. This info is used to be displayed by
the conan info command and possibly other search and report tools.
class HelloConan(ConanFile):
name = "Hello"
version = "0.1"
license = "MIT"
This attribute can contain several, comma separated licenses. It is a text string, so it can contain any text, including hyperlinks to license files elsewhere.
However, we strongly recommend packagers of Open-Source projects to use [SPDX](https://spdx.org/) identifiers from the [SPDX license list](https://spdx.org/licenses/) instead of free-formed text. This will help people wanting to automate license compatibility checks, like consumers of your package, or you if your package has Open-Source dependencies.
This is a recommended, but not mandatory attribute.
topics
Topics provide a useful way to group related tags together and to quickly tell developers what a package is about. Topics also make it easier for customers to find your recipe. It could be useful to filter packages by topics or to reuse them in Bintray package page.
The topics
attribute should be a tuple with the needed topics inside.
class ProtocInstallerConan(ConanFile):
name = "protoc_installer"
version = "0.1"
topics = ("protocol-buffers", "protocol-compiler", "serialization", "rpc")
This is an optional attribute.
user, channel
These fields are optional in a Conan reference, they could be useful to identify a forked recipe
from the community with changes specific for your company. Using these fields you may keep the
same name
and version
and use the user/channel
to disambiguate your recipe.
The value of these fields can be accessed from within a conanfile.py
:
from conans import ConanFile
class HelloConan(ConanFile):
name = "Hello"
version = "0.1"
def requirements(self):
self.requires("common-lib/version")
if self.user and self.channel:
# If the recipe is using them, I want to consume my fork.
self.requires("Say/0.1@%s/%s" % (self.user, self.channel))
else:
# otherwise, I'll consume the community one
self.requires("Say/0.1")
Only packages that have already been exported (packages in the local cache or in a remote server)
can have a user/channel assigned. For package recipes working in the user space, there is no
current user/channel by default, although they can be defined at conan install
time with:
$ conan install <path to conanfile.py> user/channel
See also
FAQ: Is there any recommendation regarding which <user> or <channel> to use in a reference?
Warning
Environment variables CONAN_USERNAME
and CONAN_CHANNEL
that were used to assign a value
to these fields are now deprecated and will be removed in Conan 2.0. Don’t use them to
populate the value of self.user
and self.channel
.
default_user, default_channel
For package recipes working in the user space, with local methods like conan install . and conan build .,
there is no current user/channel. If you are accessing to self.user
or self.channel
in your recipe,
you need to declare the environment variables CONAN_USERNAME
and CONAN_CHANNEL
or you can set the attributes
default_user
and default_channel
. You can also use python @property
:
from conans import ConanFile
class HelloConan(ConanFile):
name = "Hello"
version = "0.1"
default_user = "myuser"
@property
def default_channel(self):
return "mydefaultchannel"
def requirements(self):
self.requires("Pkg/0.1@%s/%s" % (self.user, self.channel))
settings
There are several things that can potentially affect a package being created, i.e. the final package will be different (a different binary, for example), if some input is different.
Development project-wide variables, like the compiler, its version, or the OS itself. These variables have to be defined, and they cannot have a default value listed in the conanfile, as it would not make sense.
It is obvious that changing the OS produces a different binary in most cases. Changing the compiler or compiler version changes the binary too, which might have a compatible ABI or not, but the package will be different in any case.
For these reasons, the most common convention among Conan recipes is to distinguish binaries by the following four settings, which is reflected in the conanfile.py template used in the conan new command:
settings = "os", "compiler", "build_type", "arch"
When Conan generates a compiled binary for a package with a given combination of the settings above, it generates a unique ID for that binary by hashing the current values of these settings.
But what happens for example to header only libraries? The final package for such libraries is not binary and, in most cases it will be identical, unless it is automatically generating code. We can indicate that in the conanfile:
from conans import ConanFile
class HelloConan(ConanFile):
name = "Hello"
version = "0.1"
# We can just omit the settings attribute too
settings = None
def build(self):
#empty too, nothing to build in header only
You can restrict existing settings and accepted values as well, by redeclaring the settings attribute:
class HelloConan(ConanFile):
settings = {"os": ["Windows"],
"compiler": {"Visual Studio": {"version": [11, 12]}},
"arch": None}
In this example we have just defined that this package only works in Windows, with VS 10 and 11. Any attempt to build it in other platforms with other settings will throw an error saying so. We have also defined that the runtime (the MD and MT flags of VS) is irrelevant for us (maybe we using a universal one?). Using None as a value means, maintain the original values in order to avoid re-typing them. Then, “arch”: None is totally equivalent to “arch”: [“x86”, “x86_64”, “arm”] Check the reference or your ~/.conan/settings.yml file.
As re-defining the whole settings attribute can be tedious, it is sometimes much simpler to
remove or tune specific fields in the configure()
method. For example, if our package is runtime
independent in VS, we can just remove that setting field:
settings = "os", "compiler", "build_type", "arch"
def configure(self):
self.settings.compiler["Visual Studio"].remove("runtime")
It is possible to check the settings to implement conditional logic, with attribute syntax:
def build(self):
if self.settings.os == "Windows" and self.settings.compiler.version == "15":
# do some special build commands
elif self.settings.arch == "x86_64":
# Other different commands
Those comparisons do content checking, for example if you do a typo like self.settings.os == "Windos"
,
conan will fail and tell you that is not a valid settings.os
value, and the possible range of values.
Likewise, if you try to access some setting that doesn’t exist, like self.settings.compiler.libcxx
for the Visual Studio
setting, conan will fail telling that libcxx
does not exist for that compiler.
If you want to do a safe check of settings values, you could use the get_safe()
method:
def build(self):
# Will be None if doesn't exist
arch = self.settings.get_safe("arch")
# Will be None if doesn't exist
compiler_version = self.settings.get_safe("compiler.version")
The get_safe()
method will return None
if that setting or subsetting doesn’t exist.
options
Conan packages recipes can generate different binary packages when different settings are used, but can also customize, per-package any other configuration that will produce a different binary.
A typical option would be being shared or static for a certain library. Note that this is optional, different packages can have this option, or not (like header-only packages), and different packages can have different values for this option, as opposed to settings, which typically have the same values for all packages being installed (though this can be controlled too, defining different settings for specific packages)
Options are defined in package recipes as dictionaries of name and allowed values:
class MyPkg(ConanFile):
...
options = {"shared": [True, False]}
Options are defined as a python dictionary inside the ConanFile
where each key must be a
string with the identifier of the option and the value be a list with all the possible option
values:
class MyPkg(ConanFile):
...
options = {"shared": [True, False],
"option1": ["value1", "value2"],}
Values for each option can be typed or plain strings, and there is a special value, ANY
, for
options that can take any value.
The attribute default_options
has the purpose of defining the default values for the options
if the consumer (consuming recipe, project, profile or the user through the command line) does
not define them. It is worth noticing that an uninitialized option will get the value None
and it will be a valid value if its contained in the list of valid values. This attribute
should be defined as a python dictionary too, although other definitions could be valid for
legacy reasons.
class MyPkg(ConanFile):
...
options = {"shared": [True, False],
"option1": ["value1", "value2"],
"option2": "ANY"}
default_options = {"shared": True,
"option1": "value1",
"option2": 42}
def build(self):
shared = "-DBUILD_SHARED_LIBS=ON" if self.options.shared else ""
cmake = CMake(self)
self.run("cmake . %s %s" % (cmake.command_line, shared))
...
Tip
You can inspect available package options reading the package recipe, which can be done with the command conan inspect MyPkg/0.1@user/channel.
Options
"shared": [True, False]
and"fPIC": [True, False]
are automatically managed in CMake and AutoToolsBuildEnvironment (configure/make) build helpers.
As we mentioned before, values for options in a recipe can be defined using different ways, let’s
go over all of them for the example recipe MyPkg
defined above:
Using the attribute
default_options
in the recipe itself.In the
default_options
of a recipe that requires this one: the values defined here will override the default ones in the recipe.class OtherPkg(ConanFile): requires = "MyPkg/0.1@user/channel" default_options = {"MyPkg:shared": False}
Of course, this will work in the same way working with a conanfile.txt:
[requires] MyPkg/0.1@user/channel [options] MyPkg:shared=False
It is also possible to define default values for the options of a recipe using profiles. They will apply whenever that recipe is used:
# file "myprofile" # use it as $ conan install -pr=myprofile [settings] setting=value [options] MyPkg:shared=False
Last way of defining values for options, with the highest priority over them all, is to pass these values using the command argument -o in the command line:
$ conan install . -o MyPkg:shared=True -o OtherPkg:option=value
Values for options can be also conditionally assigned (or even deleted) in the methods
configure()
and config_options()
, the
corresponding section has examples documenting these
use cases. However, conditionally assigning values to options can have it drawbacks as it is
explained in the mastering section.
One important notice is how these options values are evaluated and how the different conditionals
that we can implement in Python will behave. As seen before, values for options can be defined
in Python code (assigning a dictionary to default_options
) or through strings (using a
conanfile.txt
, a profile file, or through the command line). In order to provide a
consistent implementation take into account these considerations:
Evaluation for the typed value and the string one is the same, so all these inputs would behave the same:
default_options = {"shared": True, "option": None}
default_options = {"shared": "True", "option": "None"}
MyPkg:shared=True
,MyPkg:option=None
on profiles, command line or conanfile.txt
Implicit conversion to boolean is case insensitive, so the expression
bool(self.options.option)
:equals
True
for the valuesTrue
,"True"
and"true"
, and any other value that would be evaluated the same way in Python code.equals
False
for the valuesFalse
,"False"
and"false"
, also for the empty string and for0
and"0"
as expected.
Comparison using
is
is always equals toFalse
because the types would be different as the option value is encapsulated inside a Conan class.Explicit comparisons with the
==
symbol are case sensitive, so:self.options.option = "False"
satisfiesassert self.options.option == False
,assert self.options.option == "False"
, butassert self.options.option != "false"
.
A different behavior has
self.options.option = None
, becauseassert self.options.option != None
.
default_options
As you have seen in the examples above, recipe’s default options are declared as a dictionary with the initial desired value of the options. However, you can also specify default option values of the required dependencies:
class OtherPkg(ConanFile):
requires = "Pkg/0.1@user/channel"
default_options = {"Pkg:pkg_option": "value"}
And it also works with default option values of conditional required dependencies:
class OtherPkg(ConanFile):
default_options = {"Pkg:pkg_option": "value"}
def requirements(self):
if self.settings.os != "Windows":
self.requires("Pkg/0.1@user/channel")
For this example running in Windows, the default_options for the Pkg/0.1@user/channel will be ignored, they will only be used on every other OS.
You can also set the options conditionally to a final value with config_options()
instead of using default_options
:
class OtherPkg(ConanFile):
settings = "os", "arch", "compiler", "build_type"
options = {"some_option": [True, False]}
# Do NOT declare 'default_options', use 'config_options()'
def config_options(self):
if self.options.some_option == None:
if self.settings.os == 'Android':
self.options.some_option = True
else:
self.options.some_option = False
Important
Setting options conditionally without a default value works only to define a default value if not defined in command line. However, doing it this way will assign a final value to the option and not an initial one, so those option values will not be overridable from downstream dependent packages.
Important
Default options can be specified as a dictionary only for Conan version >= 1.8.
See also
Read more about the config_options() method.
requires
Specify package dependencies as a list or tuple of other packages:
class MyLibConan(ConanFile):
requires = "Hello/1.0@user/stable", "OtherLib/2.1@otheruser/testing"
You can specify further information about the package requirements:
class MyLibConan(ConanFile):
requires = [("Hello/0.1@user/testing"),
("Say/0.2@dummy/stable", "override"),
("Bye/2.1@coder/beta", "private")]
class MyLibConan(ConanFile):
requires = (("Hello/1.0@user/stable", "private"), )
Requirements can be complemented by 2 different parameters:
private: a dependency can be declared as private if it is going to be fully embedded and hidden from consumers of the package. It might be necessary in some extreme cases, like having to use two different versions of the same library (provided that they are totally hidden in a shared library, for example), but it is mostly discouraged otherwise.
override: packages can define overrides of their dependencies, if they require the definition of specific versions of the upstream required libraries, but not necessarily direct dependencies. For example, a package can depend on A(v1.0), which in turn could conditionally depend on Zlib(v2), depending on whether the compression is enabled or not. Now, if you want to force the usage of Zlib(v3) you can:
class HelloConan(ConanFile):
requires = ("A/1.0@user/stable", ("Zlib/3.0@other/beta", "override"))
This will not introduce a new dependency, it will just change Zlib v2 to v3 if A actually requires it. Otherwise Zlib will not be a dependency of your package.
Note
To prevent accidental override of transitive dependencies, check the config variable general.error_on_override or the environment variable CONAN_ERROR_ON_OVERRIDE.
version ranges
The syntax is using brackets:
class HelloConan(ConanFile):
requires = "Pkg/[>1.0 <1.8]@user/stable"
Expressions are those defined and implemented by [python node-semver](https://pypi.org/project/node-semver/). Accepted expressions would be:
>1.1 <2.1 # In such range
2.8 # equivalent to =2.8
~=3.0 # compatible, according to semver
>1.1 || 0.8 # conditions can be OR'ed
Go to Mastering/Version Ranges if you want to learn more about version ranges.
build_requires
Build requirements are requirements that are only installed and used when the package is built from sources. If there is an existing pre-compiled binary, then the build requirements for this package will not be retrieved.
They can be specified as a comma separated tuple in the package recipe:
class MyPkg(ConanFile):
build_requires = "ToolA/0.2@user/testing", "ToolB/0.2@user/testing"
Read more: Build requirements
exports
This optional attribute declares the set of files that should be exported and stored side by side with the conanfile.py file to make the recipe work: other python files that the recipe will import, some text file with data to read,…
The exports
field can declare one single file or pattern, or a list of any of the previous
elements. Patterns use fnmatch
formatting to declare files to include or exclude.
For example, if we have some python code that we want the recipe to use in a helpers.py
file,
and have some text file info.txt we want to read and display during the recipe evaluation
we would do something like:
exports = "helpers.py", "info.txt"
Exclude patterns are also possible, with the !
prefix:
exports = "*.py", "!*tmp.py"
exports_sources
This optional attribute declares the set of files that should be exported together with the
recipe and will be available to generate the package. Unlike exports
attribute, these files
shouldn’t be used by the conanfile.py Python code, but to compile the library or generate
the final package. And, due to its purpose, these files will only be retrieved if requested
binaries are not available or the user forces Conan to compile from sources.
The exports_sources
attribute can declare one single file or pattern, or a list of any of the
previous elements. Patterns use fnmatch
formatting to declare files to include or exclude.
Together with the source()
and imports()
methods, and the SCM feature,
this is another way to retrieve the sources to create a package. Unlike the other methods, files
declared in exports_sources
will be exported together with the conanfile.py recipe, so,
if nothing else is required, it can create a self-contained package with all the sources
(like a snapshot) that will be used to generate the final artifacts.
Some examples for this attribute are:
exports_sources = "include*", "src*"
Exclude patterns are also possible, with the !
prefix:
exports_sources = "include*", "src*", "!src/build/*"
generators
Generators specify which is the output of the install
command in your project folder. By default, a conanbuildinfo.txt file is
generated, but you can specify different generators and even use more than one.
class MyLibConan(ConanFile):
generators = "cmake", "gcc"
Check the full generators list.
should_configure, should_build, should_install, should_test
Read only variables defaulted to True
.
These variables allow you to control the build stages of a recipe during a conan build command with the optional arguments
--configure/--build/--install/--test. For example, consider this build()
method:
def build(self):
cmake = CMake(self)
cmake.configure()
cmake.build()
cmake.install()
cmake.test()
If nothing is specified, all four methods will be called. But using command line arguments, this can be changed:
$ conan build . --configure # only run cmake.configure(). Other methods will do nothing
$ conan build . --build # only run cmake.build(). Other methods will do nothing
$ conan build . --install # only run cmake.install(). Other methods will do nothing
$ conan build . --test # only run cmake.test(). Other methods will do nothing
# They can be combined
$ conan build . -c -b # run cmake.configure() + cmake.build(), but not cmake.install() nor cmake.test()
Autotools and Meson helpers already implement the same functionality. For other build systems, you can use these variables in the
build()
method:
def build(self):
if self.should_configure:
# Run my configure stage
if self.should_build:
# Run my build stage
if self.should_install: # If my build has install, otherwise use package()
# Run my install stage
if self.should_test:
# Run my test stage
Note that the should_configure
, should_build
, should_install
, should_test
variables will always be True
while building in
the cache and can be only modified for the local flow with conan build.
build_policy
With the build_policy
attribute the package creator can change the default conan’s build behavior.
The allowed build_policy
values are:
missing
: If no binary package is found, Conan will build it without the need to invoke conan install --build missing option.always
: The package will be built always, retrieving each time the source code executing the “source” method.
class PocoTimerConan(ConanFile):
build_policy = "always" # "missing"
short_paths
This attribute is specific to Windows, and ignored on other operating systems. It tells Conan to workaround the limitation of 260 chars in Windows paths.
Important
Since Windows 10 (ver. 10.0.14393), it is possible to enable long paths at the system level.
Latest python 2.x and 3.x installers enable this by default. With the path limit removed both on the OS
and on Python, the short_paths
functionality becomes unnecessary, and may be disabled explicitly
through the CONAN_USER_HOME_SHORT
environment variable.
Enabling short paths management will “link” the source
and build
directories of the package to the drive root, something like
C:\.conan\tmpdir
. All the folder layout in the local cache is maintained.
Set short_paths=True
in your conanfile.py:
from conans import ConanFile
class ConanFileTest(ConanFile):
...
short_paths = True
See also
There is an environment variable CONAN_USE_ALWAYS_SHORT_PATHS
to force
activate this behavior for all packages.
no_copy_source
The attribute no_copy_source
tells the recipe that the source code will not be copied from the source
folder to the build
folder.
This is mostly an optimization for packages with large source codebases, to avoid extra copies. It is mandatory that the source code must not be modified at all by the configure or build scripts, as the source code will be shared among all builds.
To be able to use it, the package recipe can access the self.source_folder
attribute, which will point to the build
folder when no_copy_source=False
or not defined, and will point to the source
folder when no_copy_source=True
When this attribute is set to True, the self.copy()
lines will be called twice, one copying from the source
folder and the other copying from the build
folder.
source_folder
The folder in which the source code lives.
When a package is built in the Conan local cache its value is the same as the build
folder by default. This is due to the fact that the
source code is copied from the source
folder to the build
folder to ensure isolation and avoiding modifications of shared common
source code among builds for different configurations. Only when no_copy_source=True
this folder will actually point to the package
source
folder in the local cache.
When executing Conan commands in the Package development flow like conan source, this attribute will be pointing to the folder specified in the command line.
install_folder
The folder in which the installation of packages outputs the generator files with the information of dependencies.
By default in the the local cache its value is the same as self.build_folder
one.
When executing Conan commands in the Package development flow like conan install or conan build, this attribute will be pointing to the folder specified in the command line.
build_folder
The folder used to build the source code. In the local cache a build folder is created with the name of the package ID that will be built.
When executing Conan commands in the Package development flow like conan build, this attribute will be pointing to the folder specified in the command line.
package_folder
The folder to copy the final artifacts for the binary package. In the local cache a package folder is created for every different package ID.
When executing Conan commands in the Package development flow like conan package, this attribute will be pointing to the folder specified in the command line.
cpp_info
Important
This attribute is only defined inside package_info()
method being None elsewhere.
The self.cpp_info
is responsible for storing all the information needed by consumers of a package: include directories, library names,
library paths… There are some default values that will be applied automatically if not indicated otherwise.
This object should be filled in package_info()
method.
NAME |
DESCRIPTION |
---|---|
self.cpp_info.includedirs |
Ordered list with include paths. Defaulted to |
self.cpp_info.libdirs |
Ordered list with lib paths. Defaulted to |
self.cpp_info.resdirs |
Ordered list of resource (data) paths. Defaulted to |
self.cpp_info.bindirs |
Ordered list with paths to binaries (executables, dynamic libraries,…). Defaulted to |
self.cpp_info.builddirs |
Ordered list with build scripts directory paths. Defaulted to
[""] (Package folder directory)CMake generators will search in these dirs for files like findXXX.cmake
|
self.cpp_info.libs |
Ordered list with the library names, Defaulted to |
self.cpp_info.defines |
Preprocessor definitions. Defaulted to |
self.cpp_info.cflags |
Ordered list with pure C flags. Defaulted to |
self.cpp_info.cppflags |
[DEPRECATED: use cxxflags instead] |
self.cpp_info.cxxflags |
Ordered list with C++ flags. Defaulted to |
self.cpp_info.sharedlinkflags |
Ordered list with linker flags (shared libs). Defaulted to |
self.cpp_info.exelinkflags |
Ordered list with linker flags (executables). Defaulted to |
self.cpp_info.frameworks |
Ordered list with the framework names (OSX), Defaulted to |
self.cpp_info.frameworkdirs |
Ordered list with frameworks search paths (OSX). Defaulted to |
self.cpp_info.rootpath |
Filled with the root directory of the package, see |
self.cpp_info.name |
Alternative name for the package used by generators to create files or variables.
Defaulted to the package name. Supported by cmake, cmake_multi, cmake_find_package,
cmake_find_package_multi, cmake_paths and pkg_config generators.
|
self.cpp_info.names[“generator”] |
Alternative name for the package used by an specific generator to create files or variables.
If set for a generator it will overrite the information provided by self.cpp_info.name.
Like the cpp_info.name, this is only supported by cmake, cmake_multi, cmake_find_package,
cmake_find_package_multi, cmake_paths and pkg_config generators.
|
self.cpp_info.system_libs |
Ordered list with the system library names. Defaulted to |
self.cpp_info.build_modules |
List of relative paths to build system related utility module files created by the package. Used by
CMake generators to export .cmake files with functions for consumers. Defaulted to
[] (empty) |
The paths of the directories in the directory variables indicated above are relative to the self.package_folder directory.
See also
Read package_info() for more info.
deps_cpp_info
Contains the cpp_info
object of the requirements of the recipe. In addition of the above fields, there are also properties to obtain the
absolute paths:
NAME |
DESCRIPTION |
---|---|
self.cpp_info.include_paths |
Same as |
self.cpp_info.lib_paths |
Same as |
self.cpp_info.bin_paths |
Same as |
self.cpp_info.build_paths |
Same as |
self.cpp_info.res_paths |
Same as |
self.cpp_info.framework_paths |
Same as |
self.cpp_info.build_modules_paths |
Same as |
To get a list of all the dependency names from `deps_cpp_info`
, you can call the deps member:
class PocoTimerConan(ConanFile):
...
def build(self):
# deps is a list of package names: ["Poco", "zlib", "OpenSSL"]
deps = self.deps_cpp_info.deps
It can be used to get information about the dependencies, like used compilation flags or the root folder of the package:
class PocoTimerConan(ConanFile):
...
requires = "zlib/1.2.11@conan/stable", "OpenSSL/1.0.2l@conan/stable"
...
def build(self):
# Get the directory where zlib package is installed
self.deps_cpp_info["zlib"].rootpath
# Get the absolute paths to zlib include directories (list)
self.deps_cpp_info["zlib"].include_paths
# Get the sharedlinkflags property from OpenSSL package
self.deps_cpp_info["OpenSSL"].sharedlinkflags
env_info
This attribute is only defined inside package_info()
method, being None elsewhere, so please use it only inside this method.
The self.env_info
object can be filled with the environment variables to be declared in the packages reusing the recipe.
See also
Read package_info() method docs for more info.
deps_env_info
You can access to the declared environment variables of the requirements of the recipe.
Note: The environment variables declared in the requirements of a recipe are automatically applied
and it can be accessed with the python os.environ
dictionary. Nevertheless if
you want to access to the variable declared by some specific requirement you can use the self.deps_env_info
object.
import os
class RecipeConan(ConanFile):
...
requires = "package1/1.0@conan/stable", "package2/1.2@conan/stable"
...
def build(self):
# Get the SOMEVAR environment variable declared in the "package1"
self.deps_env_info["package1"].SOMEVAR
# Access to the environment variables globally
os.environ["SOMEVAR"]
user_info
This attribute is only defined inside package_info()
method, being None elsewhere, so please use it only inside this method.
The self.user_info
object can be filled with any custom variable to be accessed in the packages reusing the recipe.
See also
Read package_info() method docs for more info.
deps_user_info
You can access the declared user_info.XXX
variables of the requirements through the self.deps_user_info
object like this:
import os
class RecipeConan(ConanFile):
...
requires = "package1/1.0@conan/stable"
...
def build(self):
self.deps_user_info["package1"].SOMEVAR
info
Object used to control the unique ID for a package. Check the package_id() to see the details of the self.info
object.
apply_env
When True
(Default), the values from self.deps_env_info
(corresponding to the declared env_info
in the requires
and build_requires
)
will be automatically applied to the os.environ
.
Disable it setting apply_env
to False if you want to control by yourself the environment variables
applied to your recipes.
You can apply manually the environment variables from the requires and build_requires:
import os
from conans import tools
class RecipeConan(ConanFile):
apply_env = False
def build(self):
with tools.environment_append(self.env):
# The same if we specified apply_env = True
pass
in_local_cache
A boolean attribute useful for conditional logic to apply in user folders local commands. It will return True if the conanfile resides in the local cache ( we are installing the package) and False if we are running the conanfile in a user folder (local Conan commands).
import os
class RecipeConan(ConanFile):
...
def build(self):
if self.in_local_cache:
# we are installing the package
else:
# we are building the package in a local directory
develop
A boolean attribute useful for conditional logic. It will be True
if the package is created with conan create, or if the
conanfile.py is in user space:
class RecipeConan(ConanFile):
def build(self):
if self.develop:
self.output.info("Develop mode")
It can be used for conditional logic in other methods too, like requirements()
, package()
, etc.
This recipe will output “Develop mode” if:
$ conan create . user/testing
# or
$ mkdir build && cd build && conan install ..
$ conan build ..
But it will not output that when it is a transitive requirement or installed with conan install.
keep_imports
Just before the build()
method is executed, if the conanfile has an imports()
method, it is
executed into the build folder, to copy binaries from dependencies that might be necessary for
the build()
method to work. After the method finishes, those copied (imported) files are removed,
so they are not later unnecessarily repackaged.
This behavior can be avoided declaring the keep_imports=True
attribute. This can be useful, for example
to repackage artifacts
scm
Warning
This is an experimental feature subject to breaking changes in future releases.
Used to clone/checkout a repository. It is a dictionary with the following possible values:
from conans import ConanFile, CMake, tools
class HelloConan(ConanFile):
scm = {
"type": "git",
"subfolder": "hello",
"url": "https://github.com/conan-io/hello.git",
"revision": "master"
}
...
type (Required): Currently only
git
andsvn
are supported. Others can be added eventually.url (Required): URL of the remote or
auto
to capture the remote from the local working copy (credentials will be removed from it). When type issvn
it can contain the peg_revision.revision (Required): id of the revision or
auto
to capture the current working copy one. When type isgit
, it can also be the branch name or a tag.subfolder (Optional, Defaulted to
.
): A subfolder where the repository will be cloned.username (Optional, Defaulted to
None
): When present, it will be used as the login to authenticate with the remote.password (Optional, Defaulted to
None
): When present, it will be used as the password to authenticate with the remote.verify_ssl (Optional, Defaulted to
True
): Verify SSL certificate of the specified url.shallow (Optional, Defaulted to
True
): Use shallow clone for Git repositories.- submodule (Optional, Defaulted to
None
): shallow
: Will sync the git submodules usingsubmodule sync
recursive
: Will sync the git submodules usingsubmodule sync --recursive
- submodule (Optional, Defaulted to
SCM attributes are evaluated in the workspace context where the conanfile.py is located before exporting it to the Conan cache, so these values can be returned from arbitrary functions that depend on the workspace layout. Nevertheless, all the other code in the recipe must be able to run in the export folder inside the cache, where it has access only to the files exported (see attribute exports) and to any other functionality from a python_requires.
Note
In case of git, by default conan will try to perform shallow clone of the repository, and will fallback to the full clone in case shallow fails (e.g. not supported by the server).
To know more about the usage of scm
check:
revision_mode
Warning
This attribute is part of the package revisions feature, so it is also an experimental feature subject to breaking changes in future releases.
This attribute allow each recipe to declare how the revision for the recipe itself should be computed. It can take three different values:
"hash"
(by default): Conan will use the checksum hash of the recipe manifest to compute the revision for the recipe.
"scm"
: the commit ID will be used as the recipe revision if it belongs to a known repository system (Git or SVN). If there is no repository it will raise an error.
python_requires
Warning
This attribute is part of the python requires feature, so it is also an experimental feature subject to breaking changes in future releases.
Python requires are associated with the ConanFile
declared in the recipe file, data
from those imported recipes is accessible using the python_requires
attribute in
the recipe itself. This attribute is a dictionary where the key is the name of the
python requires reference and the value is a dictionary with the following information:
ref
: full reference of the python requires.
exports_folder
: directory in the cache where the exported files are located.
exports_sources_folder
: directory in the cache where the files exported using theexports_sources
attribute of the python requires recipe are located.
You can use this information to copy files accompanying a python requires to the consumer workspace.:
from conans import ConanFile
class PyReq(ConanFile):
name = "pyreq"
exports_sources = "CMakeLists.txt"
def source(self):
pyreq = self.python_requires['pyreq']
self.copy("CMakeLists.txt", src=pyreq.exports_sources_folder, dst=self.source_folder)
conan_data
This attribute is a dictionary with the keys and values provided in a conandata.yml file format placed next to the conanfile.py. This YAML file is automatically exported with the recipe and automatically loaded with it too.
You can declare information in the conandata.yml file and then access it inside any of the methods of the recipe. For example, a conandata.yml with information about sources that looks like this:
sources:
"1.1.0":
url: "https://www.url.org/source/mylib-1.0.0.tar.gz"
sha256: "8c48baf3babe0d505d16cfc0cf272589c66d3624264098213db0fb00034728e9"
"1.1.1":
url: "https://www.url.org/source/mylib-1.0.1.tar.gz"
sha256: "15b6393c20030aab02c8e2fe0243cb1d1d18062f6c095d67bca91871dc7f324a"
def source(self):
tools.get(**self.conan_data["sources"][self.version])