conan install¶
$ conan install -h
Migration: Successfully updated cppstd_compat.py
usage: conan install [-h] [-v [V]] [-cc CORE_CONF] [-f FORMAT] [--name NAME]
[--version VERSION] [--user USER] [--channel CHANNEL]
[--requires REQUIRES] [--tool-requires TOOL_REQUIRES]
[-b BUILD] [-r REMOTE | -nr] [-u [UPDATE]] [-pr PROFILE]
[-pr:b PROFILE_BUILD] [-pr:h PROFILE_HOST]
[-pr:a PROFILE_ALL] [-o OPTIONS] [-o:b OPTIONS_BUILD]
[-o:h OPTIONS_HOST] [-o:a OPTIONS_ALL] [-s SETTINGS]
[-s:b SETTINGS_BUILD] [-s:h SETTINGS_HOST]
[-s:a SETTINGS_ALL] [-c CONF] [-c:b CONF_BUILD]
[-c:h CONF_HOST] [-c:a CONF_ALL] [-l LOCKFILE]
[--lockfile-partial] [--lockfile-out LOCKFILE_OUT]
[--lockfile-clean]
[--lockfile-overrides LOCKFILE_OVERRIDES] [-g GENERATOR]
[-of OUTPUT_FOLDER] [-d DEPLOYER]
[--deployer-folder DEPLOYER_FOLDER]
[--deployer-package DEPLOYER_PACKAGE] [--build-require]
[path]
Install the requirements specified in a recipe (conanfile.py or conanfile.txt).
It can also be used to install packages without a conanfile, using the
--requires and --tool-requires arguments.
If any requirement is not found in the local cache, it will iterate the remotes
looking for it. When the full dependency graph is computed, and all dependencies
recipes have been found, it will look for binary packages matching the current settings.
If no binary package is found for some or several dependencies, it will error,
unless the '--build' argument is used to build it from source.
After installation of packages, the generators and deployers will be called.
positional arguments:
path Path to a folder containing a recipe (conanfile.py or
conanfile.txt) or to a recipe file. e.g.,
./my_project/conanfile.txt.
optional arguments:
-h, --help show this help message and exit
-v [V] Level of detail of the output. Valid options from less
verbose to more verbose: -vquiet, -verror, -vwarning,
-vnotice, -vstatus, -v or -vverbose, -vv or -vdebug,
-vvv or -vtrace
-cc CORE_CONF, --core-conf CORE_CONF
Define core configuration, overwriting global.conf
values. E.g.: -cc core:non_interactive=True
-f FORMAT, --format FORMAT
Select the output format: json
--name NAME Provide a package name if not specified in conanfile
--version VERSION Provide a package version if not specified in
conanfile
--user USER Provide a user if not specified in conanfile
--channel CHANNEL Provide a channel if not specified in conanfile
--requires REQUIRES Directly provide requires instead of a conanfile
--tool-requires TOOL_REQUIRES
Directly provide tool-requires instead of a conanfile
-b BUILD, --build BUILD
Optional, specify which packages to build from source.
Combining multiple '--build' options on one command
line is allowed. Possible values: --build="*" Force
build from source for all packages. --build=never
Disallow build for all packages, use binary packages
or fail if a binary package is not found, it cannot be
combined with other '--build' options. --build=missing
Build packages from source whose binary package is not
found. --build=cascade Build packages from source that
have at least one dependency being built from source.
--build=[pattern] Build packages from source whose
package reference matches the pattern. The pattern
uses 'fnmatch' style wildcards. --build=~[pattern]
Excluded packages, which will not be built from the
source, whose package reference matches the pattern.
The pattern uses 'fnmatch' style wildcards.
--build=missing:[pattern] Build from source if a
compatible binary does not exist, only for packages
matching pattern.
-r REMOTE, --remote REMOTE
Look in the specified remote or remotes server
-nr, --no-remote Do not use remote, resolve exclusively in the cache
-u [UPDATE], --update [UPDATE]
Will install newer versions and/or revisions in the
local cache for the given reference, or all in case no
argument is supplied. When using version ranges, it
will install the latest version that satisfies the
range. Also, if using revisions, it will update to the
latest revision for the resolved version range.
-pr PROFILE, --profile PROFILE
Apply the specified profile. By default, or if
specifying -pr:h (--profile:host), it applies to the
host context. Use -pr:b (--profile:build) to specify
the build context, or -pr:a (--profile:all) to specify
both contexts at once
-pr:b PROFILE_BUILD, --profile:build PROFILE_BUILD
-pr:h PROFILE_HOST, --profile:host PROFILE_HOST
-pr:a PROFILE_ALL, --profile:all PROFILE_ALL
-o OPTIONS, --options OPTIONS
Apply the specified options. By default, or if
specifying -o:h (--options:host), it applies to the
host context. Use -o:b (--options:build) to specify
the build context, or -o:a (--options:all) to specify
both contexts at once. Example: -o pkg:with_qt=true
-o:b OPTIONS_BUILD, --options:build OPTIONS_BUILD
-o:h OPTIONS_HOST, --options:host OPTIONS_HOST
-o:a OPTIONS_ALL, --options:all OPTIONS_ALL
-s SETTINGS, --settings SETTINGS
Apply the specified settings. By default, or if
specifying -s:h (--settings:host), it applies to the
host context. Use -s:b (--settings:build) to specify
the build context, or -s:a (--settings:all) to specify
both contexts at once. Example: -s compiler=gcc
-s:b SETTINGS_BUILD, --settings:build SETTINGS_BUILD
-s:h SETTINGS_HOST, --settings:host SETTINGS_HOST
-s:a SETTINGS_ALL, --settings:all SETTINGS_ALL
-c CONF, --conf CONF Apply the specified conf. By default, or if specifying
-c:h (--conf:host), it applies to the host context.
Use -c:b (--conf:build) to specify the build context,
or -c:a (--conf:all) to specify both contexts at once.
Example: -c tools.cmake.cmaketoolchain:generator=Xcode
-c:b CONF_BUILD, --conf:build CONF_BUILD
-c:h CONF_HOST, --conf:host CONF_HOST
-c:a CONF_ALL, --conf:all CONF_ALL
-l LOCKFILE, --lockfile LOCKFILE
Path to a lockfile. Use --lockfile="" to avoid
automatic use of existing 'conan.lock' file
--lockfile-partial Do not raise an error if some dependency is not found
in lockfile
--lockfile-out LOCKFILE_OUT
Filename of the updated lockfile
--lockfile-clean Remove unused entries from the lockfile
--lockfile-overrides LOCKFILE_OVERRIDES
Overwrite lockfile overrides
-g GENERATOR, --generator GENERATOR
Generators to use
-of OUTPUT_FOLDER, --output-folder OUTPUT_FOLDER
The root output folder for generated and build files
-d DEPLOYER, --deployer DEPLOYER
Deploy using the provided deployer to the output
folder. Built-in deployers: 'full_deploy',
'direct_deploy', 'runtime_deploy'
--deployer-folder DEPLOYER_FOLDER
Deployer output folder, base build folder by default
if not set
--deployer-package DEPLOYER_PACKAGE
Execute the deploy() method of the packages matching
the provided patterns
--build-require Whether the provided path is a build-require
The conan install
command is one of the main Conan commands, and it is used to resolve and install dependencies.
This command does the following:
Compute the whole dependency graph, for the current configuration defined by settings, options, profiles and configuration. It resolves version ranges, transitive dependencies, conditional requirements, etc, to build the dependency graph.
Evaluate the existence of binaries for every package in the graph, whether or not there are precompiled binaries to download, or if they should be built from sources (as directed by the
--build
argument). If binaries are missing, it will not recompute the dependency graph to try to fallback to previous versions that contain binaries for that configuration. If a certain dependency version is desired, it should be explicitly required.Download precompiled binaries, or build binaries from sources in the local cache, in the right order for the dependency graph.
Create the necessary files as requested by the “generators”, so build systems and other tools can locate the locally installed dependencies
Optionally, execute the desired
deployers
.
See also
Check the JSON format output for this command.
Conanfile path or –requires¶
The conan install
command can use 2 different origins for information. The first one is using a local conanfile.py
or conanfile.txt
, containing definitions of the dependencies and generators to be used.
$ conan install . # there is a conanfile.txt or a conanfile.py in the cwd
$ conan install conanfile.py # also works, direct reference file
$ conan install myconan.txt # explicit custom name
$ conan install myfolder # there is a conanfile in "myfolder" folder
Even if it is possible to use a custom name, in the general case, it is recommended to use the default conanfile.py
name, located in the repository root, so users can do a straightforward git clone ... `` + ``conan install .
The other possibility is to not have a conanfile
at all, and define the requirements to be installed directly in the
command line:
# Install the zlib/1.2.13 library
$ conan install --requires=zlib/1.2.13
# Install the zlib/1.2.13 and bzip2/1.0.8 libraries
$ conan install --requires=zlib/1.2.13 --requires=bzip2/1.0.8
# Install the cmake/3.23.5 and ninja/1.11.0 tools
$ conan install --tool-requires=cmake/3.23.5 --tool-requires=ninja/1.11.0
# Install the zlib/1.2.13 library and ninja/1.11.0 tool
$ conan install --requires=zlib/1.2.13 --tool-requires=ninja/1.11.0
In the general case, it is recommended to use a conanfile
instead of defining things in the command line.
Profiles, Settings, Options, Conf¶
There are several arguments that are used to define the effective profiles that will be used, both for the “build” and “host” contexts.
By default the arguments refer to the “host” context, so --settings:host, -s:h
is totally equivalent to
--settings, -s
. Also, by default, the conan install
command will use the default
profile both for the
“build” and “host” context. That means that if a profile with the “default” name has not been created, it will error.
Multiple definitions of profiles can be passed as arguments, and they will compound from left to right (right has the highest priority)
# The values of myprofile3 will have higher priority
$ conan install . -pr=myprofile1 -pr=myprofile2 -pr=myprofile3
If values for any of settings
, options
and conf
are provided in the command line, they create a profile that
is composed with the other provided -pr
(or the “default” one if not specified) profiles, with higher priority,
not matter what the order of arguments is.
# the final "host" profile will always be build_type=Debug, even if "myprofile"
# says "build_type=Release"
$ conan install . -pr=myprofile -s build_type=Debug
Generators and deployers¶
The -g
argument allows to define in the command line the different built-in generators to be used:
$ conan install --requires=zlib/1.2.13 -g CMakeDeps -g CMakeToolchain
Note that in the general case, the recommended approach is to have the generators
defined in the conanfile
,
and only for the --requires
use case, it would be more necessary as command line argument.
Generators are intended to create files for the build systems to locate the dependencies, while the deployers
main use case is to copy files from the Conan cache to user space, and performing any other custom operations over the dependency graph,
like collecting licenses, generating reports, deploying binaries to the system, etc. The syntax for deployers is:
# does a full copy of the dependencies binaries to the current user folder
$ conan install . --deployer=full_deploy
There are 3 built-in deployers:
full_deploy
does a complete copy of the dependencies binaries in the local folder, with a minimal folder structure to avoid conflicts between files and artifacts of different packagesdirect_deploy
does a copy of only the immediate direct dependencies, but does not include the transitive dependencies.runtime_deploy
deploys all the shared libraries and the executables of the dependencies (like.so
,.dll
, or.dylib
files) into a flat directory structure. (Available since Conan 2.5.0)
Some generators might have the capability of redefining the target “package folder”. That means that if some other
generator like CMakeDeps
is used that is pointing to the packages, it will be pointing to the local deployed
copy, and not to the original packages in the Conan cache. See the full example in Creating a Conan-agnostic deploy of dependencies for developer use.
It is also possible, and it is a powerful extension point, to write custom user deployers. Read more about custom deployers in Deployers.
It is possible to also invoke the package recipes deploy()
method with the --deployer-package
:
# Execute deploy() method of every recipe that defines it
$ conan install --requires=pkg/0.1 --deployer-package=*
# Execute deploy() method only for "pkg" (any version) recipes
$ conan install --requires=pkg/0.1 --deployer-package=pkg/*
The --deployer-package
argument is a pattern and accept multiple values, all package references matching any of the defined patterns will execute its deploy()
method. The --deployer-folder
argument will also affect the output location of this deployment. See the deploy() method.
If multiple deployed packages deploy to the same location, it is their responsibility to not mutually overwrite their binaries if they have the same filenames. For example if multiple packages deploy()
a file called “License.txt”, each recipe is responsible for creating an intermediate folder with the package name and/or version that makes it unique, so other recipes deploy()
method do not overwrite previously deployed “License.txt” files.
Name, version, user, channel¶
The conan install
command provides optional arguments for --name, --version, --user, --channel
. These
arguments might not be necessary in the majority of cases. Never for conanfile.txt
and for conanfile.py
only in the case that they are not defined in the recipe:
from conan import ConanFile
from conan.tools.scm import Version
class Pkg(ConanFile):
name = "mypkg"
def requirements(self):
if Version(self.version) >= "3.23":
self.requires("...")
# If we don't specify ``--version``, it will be None and it will fail
$ conan install . --version=3.24
Lockfiles¶
The conan install
command has several arguments to load and produce lockfiles.
By default, if a conan.lock
file is located beside the recipe or in the current working directory
if no path is provided, will be used as an input lockfile.
Lockfiles are strict by default, that means that
if there is some requires
and it cannot find a matching locked reference in the lockfile, it will error
and stop. For cases where it is expected that the lockfile will not be complete, as there might be new
dependencies, the --lockfile-partial
argument can be used.
By default, conan install
will not generate an output lockfile, but if the --lockfile-out
argument
is provided, pointing to a filename, like --lockfile-out=result.lock
, then a lockfile will be generated
from the current dependency graph. If --lockfile-clean
argument is provided, all versions and revisions
not used in the current dependency graph will be dropped from the resulting lockfile.
Let’s say that we already have a conan.lock
input lockfile, but we just added a new requires = "newpkg/1.0"
to a new dependency. We could resolve the dependencies, locking all the previously locked versions, while allowing
to resolve the new one, which was not previously present in the lockfile, and store it in a new location, or overwrite the existing lockfile:
# --lockfile=conan.lock is the default, not necessary
$ conan install . --lockfile=conan.lock --lockfile-partial --lockfile-out=conan.lock
Also, it is likely that the majority of lockfile operations are better managed by the conan lock
command.
See also
Read the tutorial about the local package development flow.
Update¶
The conan install
command has an --update
argument that will force the re-evaluation of the selected items of the dependency graph,
allowing for the update of the dependencies to the latest version if using version ranges, or to the latest revision of the same version,
when those versions are not locked in the given lockfile. Passing --update
will check every package in the dependency graph,
but it is also possible to pass a package name to the --update
argument (it can be added to the command more than once with different names),
to only update those packages, which avoids the re-evaluation of the whole graph.
$ conan install . --update # Update all packages in the graph
$ conan install . --update=openssl # Update only the openssl package
$ conan install . --update=openssl --update=boost # Update both openssl and boost packages
Note that the --update
argument will look into all the remotes specified in the command for possible newer versions,
and won’t stop at the first newer one found.