Generate conda environment files from Python source code
Project description
Purpose
The goal of conda_deps
is to generate a conda environment file as a result of
the dependencies found in a repository. At the moment, it only translates Python and R dependencies
but it would be great to have it working for other programming languages as well.
conda_deps
translates import statements in Python source code like:
import numpy
import scipy
into an environment file:
name: testenv
channels:
- conda-forge
- bioconda
- defaults
dependencies:
- python
- numpy
- scipy
For R it translates library imports like:
library(reshape2)
library(ggplot2)
into:
name: testenv
channels:
- conda-forge
- bioconda
- defaults
dependencies:
- r-base
- r-reshape2
- r-ggplot2
Warning
Please note that conda_deps
does not check dependencies in a clever way. For example, if your code imports scipy
and numpy
, the script will generate an environment with both listed even though numpy
is a dependency of scipy
and only the latter would be required. So the expected output of conda_deps
is a direct translation of the dependencies found in your code.
Installation
conda_deps
only works in Python 3 and will only scan properly Python 3 source code.
There should be no restriction in the case of R.
conda_deps
has been uploaded to conda-forge
so you can install it with:
# if you don't have conda available:
curl -O https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh -b -p conda-install
source conda-install/etc/profile.d/conda.sh
conda update --all --yes
# once conda is available:
conda create --name conda_deps --channel conda-forge conda_deps
conda activate conda_deps
conda_deps --help
Usage
This is how you scan a single Python or R file:
conda_deps </path/to/filename>
The script can also scan folders:
conda_deps </path/to/folder/>
In case you want to exclude one or more subfolders, use the --exclude-folder
option one or more times:
conda_deps --exclude-folder </path/to/folder/folder1> </path/to/folder>
You may also want to scan additonal files of folders:
conda_deps </path/to/folder> --include-files my-script.py --include-files </another/folder>
How it works
Python source code
The script uses Python's Abstract Syntax Trees
to parse files ending in .py
. It looks for import <module>
statements, and discards the modules belonging to the
Python Standard Library (e.g. import os
). It assumes that <module>
has a corresponding conda package
with the same name (e.g. import numpy
corresponds to conda install numpy
). However, that is not
always the case and you can provide a proper translation between the module name and its corresponding
conda package (e.g. import yaml
will require conda install pyyaml
) via the
python_deps.json file, which
will be loaded into a dictionary at the beginning of the script. It looks like this:
{
"Bio":"biopython",
"Cython":"cython",
"bs4":"beautifulsoup4",
"bx":"bx-python",
"lzo":"python-lzo",
"pyBigWig":"pybigwig",
"sklearn":"scikit-learn",
"web":"web.py",
"weblogolib":"python-weblogo",
"yaml":"pyyaml"
}
The dictionary key is the name in import <module>
and the value is the name of the conda package.
The python_deps.json file is meant to be useful for generic use. However, it is possible to include additional json files specific to your project:
conda_deps --include-py-json my_project.json </path/to/project/>
The translations in my_project.json will take priority over those in python_deps.json.
If you find that there are missing translations in the general purpose python_deps.json file, please feel free to open a pull request to add more.
R source code
In the case of R files, it uses grep
to look for library(name)
regular expressions in files ending in .R
.
The same way we use a json
file to detail translations for Python,
we use the r_deps.json
file which will be loaded into a dictionary at the beginning of the script. Here is how it looks like:
{
"dplyr":"r-dplyr",
"edgeR":"bioconductor-edger",
"flashClust":"r-flashclust",
"gcrma":"bioconductor-gcrma",
"ggplot2":"r-ggplot2",
"gplots":"r-gplots",
"gridExtra":"r-gridextra",
"grid":"r-gridbase",
"gtools":"r-gtools",
"hpar":"bioconductor-hpar",
"knitr":"r-knitr",
"limma":"bioconductor-limma",
"maSigPro":"bioconductor-masigpro",
}
In this case the dictionary key is the name in library(name)
and the value is the name of the conda package.
If you are missing a translation in r_deps.json you can either open a pull request to add it or include it in your own json file:
conda_deps --include-r-json my_project.json </path/to/project/>
Please note that the translations in my_project.json will take priority over those in r_deps.json.
Warning
An important point to bear in mind is that the translations for both Python and R are not comprehensive and are mainly based in the dependencies used in the past. It will be a matter of time to keep adding new dependencies to the json files in charge of the translation. This implies that the environment file produced as output may not be valid straight away and conda will complain about that when creating the environment (i.e. error message: PackagesNotFoundError).
Related tools
- snakefood: a more comprehensive tool but it works only with Python 2.
- pipreqs: does a similar job but for requirements.txt files and pip.
References
- https://docs.python.org/3/library/ast.html#module-ast
- http://bit.ly/2rDf5xu
- http://bit.ly/2r0Uv9t
- https://github.com/titusjan/astviewer
Changelog
- v0.0.9:
- Scan .Rmd and .ipynb files as well, therefore the script now depends on nbconvert
- Able to scan Python imports with multiple modules (e.g.
import numpy, matplotlib
) - Add new R dependencies to the json dictionary
- Scan
rmagic
andcythonmagic
in .ipynb files (#1)
- v0.0.8:
- v0.0.7:
- Improve the test to check whether a module belongs to the Python Standard Library
- Add new R dependencies to the json dictionary
- v0.0.6:
- Add sanity check for dependency translation
- Improve regex to translate Bioconductor dependencies
- v0.0.5:
- Add r_deps.json to manifest file
- Rename option --include-py-files to --include-files
- v0.0.4:
- Add translation of R dependencies
- Not uploaded to conda-forge due to missing r_deps.json in the manifest file
- v0.0.3: minor bugfixes.
- v0.0.2: first working version uploaded to conda-forge.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
File details
Details for the file conda_deps-0.0.9.tar.gz
.
File metadata
- Download URL: conda_deps-0.0.9.tar.gz
- Upload date:
- Size: 12.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/1.13.0 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.0.1 requests-toolbelt/0.9.1 tqdm/4.32.1 CPython/3.7.3
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | d12567fe9719befaa571c6167fff043338552399c5076910e5022e12bfd02c53 |
|
MD5 | 31a4926c9bd62cbcb3b305dc97875a9f |
|
BLAKE2b-256 | 21659f474da081023749a4d906000069cb280ced32dec8d6d43791818d06b7cb |