https://github.com/miraculixx/runfast

runfast speeds up PyCharm's package scanning by 90%
https://github.com/miraculixx/runfast

conda performance pycharm

Last synced: 8 days ago
JSON representation

runfast speeds up PyCharm's package scanning by 90%

• Host: GitHub
• URL: https://github.com/miraculixx/runfast
• Owner: miraculixx
• Created: 2022-02-08T23:33:17.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2024-05-16T17:55:38.000Z (6 months ago)
• Last Synced: 2024-10-07T20:02:35.108Z (29 days ago)
• Topics: conda, performance, pycharm
• Language: Python
• Homepage:
• Size: 3.91 KB
• Stars: 1
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README

Awesome Lists containing this project

README

        
Avoid PyCharm Overloading

=========================

This speeds up PyCharm's package index processes and avoids CPU & memory

overloading.

On my machine (4c/32GB) with 5 conda environments and 10 projects, this reduces

PyCharm's package scanning from 15 minutes to 45 seconds.

Why?

----

When PyCharm has multiple conda environments, it will sometimes launch concurrent

package scanning across those environments. To do so, it uses the `conda` cli,

and a tool called conda_packaging_tool.py, available from the JetBrains

intellij-community github repository (see below).

Unfortunately, these tools do not behave well when launched in parallel, as

they tend to overload CPU & memory in this case.

How?

----

To avoid stalling these processes, we modify their sources as follows:

1. modify the conda cli

    $ conda activate base

    $ pip install runfast

    $ nano `which conda`

    Modify the main section:

     if __name__ == '__main__':

        from conda.cli import main

        from runfast import cached

        cached(main, on=lambda argv: argv[1] == 'list' if len(argv) > 1 else False)

2. modify the conda_packaging_tool.py

    # find $HOME -name conda_packaging_tool.py

    if __name__ == '__main__':

        from runfast import cached

        cached(main)

How does this work?

-------------------

Two simple steps:

1. When called in parallel, only one process is allowed to proceed.

2. `runfast.cached` caches the output of these tools (stdout, stderr) for 1 minute,

   given the same command line parameters

That is, when PyCharm launches 5 package scanning (`conda list`) commands, only

one of them will run immediately. If some of the scans are for the same environment,

only one of them will actually run, while the others simply return the same output.

How to clear the cache or avoid caching at all?

    $ export RUNFAST_NOCACHE=1

Testing

-------

a) conda packaging tool

    # first time

    $ time python /opt/pycharm-community-2021.2.1/plugins/python-ce/helpers/conda_packaging_tool.py

    real    0m13.271s

    user    0m9.631s

    sys     0m1.936s

    # second time

    $ time python /opt/pycharm-community-2021.2.1/plugins/python-ce/helpers/conda_packaging_tool.py

    real    0m3.159s

    user    0m0.139s

    sys     0m1.298s

b) conda cli

   # first time

   $ time conda list -p /path/to/env

   real    0m7.985s

   user    0m7.883s

   sys     0m0.087s

   # second time

   real    0m0.152s

   user    0m0.116s

   sys     0m0.032s

References

----------

* https://github.com/JetBrains/intellij-community/blob/master/python/helpers/conda_packaging_tool.py

* https://github.com/conda/conda/blob/33a142c16530fcdada6c377486f1c1a385738a96/conda/core/index.py#L53