Pip freeze for only project requirements
Solution 1:
pipreqs can save the day for a specific project. Just
pip install pipreqs
#then
pipreqs path/to/project
Github Page
Solution 2:
I have tried both pipreqs and pigar and found pigar is better because it also generates information about where it is used, it also has more options.
Solution 3:
I use this command
EDIT: Thanks Addisson Klinke for suggestion
pip freeze -r requirements.txt | grep -B100 "pip freeze" | grep -v "pip freeze"
pip freeze -r requirements.txt | sed '/freeze/,$ d'
When I ran pip freeze -r requirements.txt the output is something like
APScheduler==3.2.0
Eve==0.6.4
Eve-Elastic==0.3.8
## The following requirements were added by pip freeze:
arrow==0.8.0
Cerberus==0.9.2
I have a requirements file like this
APScheduler
Eve
Eve-Elastic
So I get this output and sed to remove the dependencies that I don`t want.
First output this to a file
pip freeze -q -r requirements.txt | sed '/freeze/,$ d' > requirements-froze.txt
That will output just the libs with version
APScheduler==3.2.0
Eve==0.6.4
Eve-Elastic==0.3.8
Then replace requirements file
mv requirements-froze.txt requirements.txt
Solution 4:
I still suggest using the official pip freeze > requirements.txt (documentation) compared to the two alternative features using pigar and pipreqs mentioned in the other answers because pip freeze lists the effective package names.
Incomplete comparison among the three as per February 2022
| Criteria \ Tool | pip freeze > requirements.txt | pigar | pipreqs |
|---|---|---|---|
| Name mismatch (1) | Package my-package==1.0.0
|
Module my_package == 1.0.0
|
Module my_package==1.0.0
|
| Module overloading (2) | All packages my-package1==1.0.0, my-package2==2.0.0
|
None | Top-level module (version shows 0.0.0) my==0.0.0
|
| Showing only directly used packages | No | Yes | Yes |
| Minimal contents | Yes | No | Yes |
(1) There can be a mismatch between module and package name such as my-package (package name) vs my_package (module name).
(2) There can be several packages using the same top level folder such as my-package1 and my-package2 (package names) which are installed under my/package1 and my/package2, which are imported by Python's command import my.package1 and import my.package2. Note that pipreqs notes version 0.0.0 for the not existing package my.
I know these are very particular cases but I hope giving people this overview can help understanding limitations and avoid possible mistakes.