You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Community feedback needed for project's development guidelines. #805
I've been participating on this project in the background whilst developing my
own views of pymode at my own fork.
I've seen some good issues and points being brought up occasionally by other
users and I think this is a good oportunity to settle on some 'rules' and
practices for this project. The idea is to provide clearer guidelines on
contributing, future directions, etc.
So now its the time to speak up.
After our conclusions on this topic I'll commit myself to be more active and
assertive in closing/accepting pull requests.
So from my part I say we must have some sort of testing before accepting PRs.
I will put a draft on this topic later on. You can check it out here:
Should be tested before accepted (open to discussion on methodology).
Python code
Should conform to PEP8 and PEP257 (open to discussion).
Important feature requests
Priority:
Improve python-mode's communication (including our one-liner slogan and our readme.rst).
Redesign our slogan (make it more IDE-like).
Redesign our readme.rst.
Move third party libraries to submodules.
Advantages:
Provide modularity for our code base.
Moves and unnecessary burden of keeping the code base of different
projects up to date.
Allow users to handpick their preferred versions of each library.
Disadvantages:
Forces users to issue and additional git command.
Cleanup libraries and huge codebase on pymode/libs.
Create a roadmap of python libraries that we want to include.
We should have a clear line on what to include and not to include,
specially because adding up libraries probably would give diminishing
returns (there is some overlap between them).
Non priority:
Find a suitable library for completion (probably jedi).
The most important changes are that libraries found in pymode/libs and other
files were moved to submodules.
Other minor improvements were also executed.
This greatly improves the maintenance in our project as thousand-lines long
commits are not necessary. To update all submodules to their latest version the
following commands must be issued:
The only problem I see with using submodules is with Vim package managers like Vundle. Right now all that is required to make python-mode work is to add a Plugin line in vimrc and run PluginInstall. With submodules, one would have to go into the cloned repo and run submodule init and update commands by hand.
I have been using python-mode, with the default configuration for a couple of weeks. Thanks for creating this.
Couple of suggestions which may make it more popular and lead to continued development and improvements:
I agree, that we should move towards simplicity, rather than ask the new user to learn more things before being able to use pymode. There are some arguments against making VIM into an IDE, but pymode, is kindof a python IDE pluggable into VIM, and thats why I find it amazing. I am guessing a lot of people, start using pymode because they miss the "integrated" development that an IDE provides, but also want to use VIM (and not the pluging for VIM provided by the IDE). Good software provides an abstraction for the underlying complexity. Further, the tagline reads "Vim python-mode. PyLint, Rope, Pydoc, breakpoints from box." Unless you have been using Python for a while, people don't usually know what these terms mean. Even though Pycharm probably uses some of these libraries, or something similar in the background, they dont say anything about these.
Why don't we advertise it as such, such as "Python IDE for VIM"?
The number of contributors to a project like this will be proportional to the number of users. With more user-friendly UIs like Pycharm, most new Python users are going to Pycharm. I wonder why we have such a lengthy first page. This would lead to the "paradox of choice" https://en.wikipedia.org/wiki/The_Paradox_of_Choice. Here's the launch page for Tensorflow https://github.com/tensorflow/tensorflow. I am pretty sure Tensorflow is way more complicated than python-mode. Maybe we can list the high-level functionalities provided by Pymode with some videos/demonstrations, and leave the rest to the wiki?
I notice that there are multiple libraries being used, some of which maybe redundant, not developed anymore or maybe there are better options available since 2015. http://liuchengxu.org/posts/use-vim-as-a-python-ide/. Can we use some of the more mordern ones?
In general, I have seen that VIM has tonnes of plugins for the same function, and it takes a while to figure out which one maybe best suited. Where python-mode could really add value for Python coders in VIM, is to screen all these plugins and provide, the best ones, great defaults, such that the user can get a IDE like environment quickly and get back to coding.
Thanks a lot for creating pymode. I can help with some of these.. or other ideas from the community over time.
tl;dr: submodules structure benefits outweighs its costs.
Those shortcomings are known but I think the benefits outweighs the costs in
this case. To update a single library usually means removing one folder and
replacing with another one full of changes that are tracked (and shouldn't be).
And this a mice and rat game because we will keep having to "paste" libraries
into our folder.
Another point is that in the makefile we are installing those libraries via pip
which I also think is not the best approach. Vim should not interfere with the
user's libraries.
And to remedy this issue one command is enough: git submodule update --init --recursive && git submodule update --recursive --remote
My idea is to move the tip of this branch to master (thus keeping the current
layout) and change develop to this new structure and keep it this way for 1-3
months to improve user testing and collect feedback. Also the libs folder is
overcomplicated now and I suspect we don't use a good part of the code that is
there.
Totally agree on this improvement of the plugin. Good point.
(Same argument as 1.) Redesign of the readme.rst is needed.
This point is hairy I think for two reasons.
First I don't think we can change the license without the author's consent
(I'm not the author).
Second I don't think this software clashes with corporate interests/IP
because you are only required to release source code if you distribute this
software as part of your product. In other words, corporations can use this
software to develop proprietary python applications and distribute them with
whatever licenses. If they distribute their product with embedded
python-mode in it then they have to distribute the source code too (which
I hardly think is the case).
I will put this in our roadmap. The libs folder is still obscure to me (in
terms of what is actually used and what is not) and mapping them is
a priority.
Great feedback! And I also thank those who created python mode :)
python-mode is a great. Thanks for your work.
There are many plugins for flake8, but submodule or sub-folder is too hard to use that.
When I register flake8 to git hooks, it will has two copies(maybe version also is different).
Use the user's library as an option?
English is not my native laguage.
If I understood correctly you also oppose the use of submodules.
Again I think that the benefits outweigh the costs as a single command has to
be issued and the user can either decide which version of each library he wants
to use.
Also I don't want vim to interfere with a users choice of python packages (pip
packages).
Flake8 sounds like a good idea. But before we start adding more
packages/complexity to our mix I think we should focus on what we have right
now.
Thus I suggest that we create a list of linters/tools and discuss on those that
should be implemented. But again our starting point should be what we have
right now and have them operating correctly.
I will work on the readme a bit over the next few days.. I have been going over the options for upgrading the libraries over the weekend a bit:
There have been a lot of interesting developments going on in the development of "Language server protocols", which might unite the need for separate libraries for separate editors and languages.. but I am not sure how far out that is: https://github.com/prabirshrestha/vim-lsp
It seems that almost universally people are using jedi instead of rope for autocompletion
In terms of linting docstrings, pep257 has be renamed/upgraded to pydocstyle.
for formatting, Google/yapf seems to be more popular than autopep8.
a lot of the complaints regarding using pymode vs IDEs is that the lack of async in VIM 7.4 makes it slow to manage large libraries. So people switch to Pycharm.
For refactoring, rope is pretty much all there is for VIM.
However, none of these are well integrated like python-mode. You have to do all of it on a one by one basis..
I'll go through each point later but something occurred to me.
Assume the following are true:
Most users of vim use it to code in more than one programming language.
Advanced users of vim are aware of its code completion features and plugins.
Then I think that the following is a conclusion: each user will setup its own
code completion tool.
Thus I think that we can in the future add a robust completion feature that
should be off by default (let's go with jedi if what you said is true) but for
now assume that the users will have their own completions set up.
The idea behind 'outsourcing' this is to let us focus on the most pressing
needs that we have right now (the readme.md, the roadmap for needed
maintenances and then adding new features IMO; adding support for jedi being
part of the third group).
In terms of modularity, we should probably try to delegate as much of the work as possible to third party plugins. That way we can reduce our work, and as and when those plugins improve, we can just pull the improvements into this one.
Before we decide for how to modify this library, so that its more useful for everybody and so that we can sustain it long term, should we come up with a list of features/specs that are necessary for a IDE-type workflow for a moderately sized project in VIM?
Perhaps, we can chose a reasonably large python project OS project (sklearn, tensorflow etc), and try to figure out whether the current plugin can meet the specs.
If not, what are the bottlenecks, and how to go about fixing that. Otherwise, we might be fixing problems that dont need to be fixed or dont add value.
There have been a lot of interesting developments going on in the development of "Language server protocols", which might unite the need for separate libraries for separate editors and languages.. but I am not sure how far out that is: https://github.com/prabirshrestha/vim-lsp
Sounds good but I don't quite grasp the tangible aspect of it.
It seems that almost universally people are using jedi instead of rope for autocompletion
Accepted; already on top post.
In terms of linting docstrings, pep257 has be renamed/upgraded to pydocstyle.
I think that the idea in the long run is to have this IDE-like
experience. However I'm not sure what the trade-offs are and I have not
tried neovim (nor oni).
And well, this is Vim, so when faced with any trade-offs between 'bells
and whistles' and vim's 'culture' so to speak we should go the vim-way.
for formatting, Google/yapf seems to be more popular than autopep8.
I have not heard about it so far but it looks just great. Added to top
post!
a lot of the complaints regarding using pymode vs IDEs is that the lack of
async in VIM 7.4 makes it slow to manage large libraries. So people switch
to Pycharm.
Vim is not for everyone. And I don't think that this is a problem, nor
a matter of 'better vs worse'. It is just a matter of style. It sure does
not mean that we rejoice in having things barebones-vim-like. We should
go after improvements and functionality but we will never be a GUI-IDE.
Let's keep this in mind.
For refactoring, rope is pretty much all there is for VIM.
Then we go with Rope ;)
Two weeks have passed since the opening of this post. I think we are able to
go with submodules, freeze this discussion for a while and work on this small
roadmap that we have agreed upon.
I haven't used it my self, but it looks like a nice way to manage 'external dependencies'.
This would add some extra effort to the development community (but not the craziness of copying files around), make the git history a little bit dirtier, but would ease the life of our users, avoiding the need of using extra git commands to have the plugin working.
I think submodules is what we are looking for. We already have 6 or 7 plugins
and this number is likely to increase in the future. As we have seen those
changes from one version to another can be several hundreds of lines long.
The conclusion that I have is that if I want to track pymode's file I'll have
a lot of difficulty.
One command can sometimes be a hindrance but I think it still wins over
subtrees.
We could try that on a different branch in the future and see how it goes :)
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
Activity
fmv1992 commentedon Nov 8, 2017
Dear community,
I'm pleased to announce commit bf018e2.
The most important changes are that libraries found in
pymode/libsand otherfiles were moved to submodules.
Other minor improvements were also executed.
This greatly improves the maintenance in our project as thousand-lines long
commits are not necessary. To update all submodules to their latest version the
following commands must be issued:
git submodule update --init --recursivegit submodule update --recursive --remoteI strongly the community to check out the new
dev_submodulesbranch and testit. If you do so do not forget to execute the following command:
git submodule update --init --recursivegit submodule update --recursive --remoteI hope to hear feedback from you soon!
Best,
Felipe.
PS: Just for the record, the 'legacy' commit is
b848f62.
MadWombat commentedon Nov 8, 2017
The only problem I see with using submodules is with Vim package managers like Vundle. Right now all that is required to make python-mode work is to add a Plugin line in vimrc and run PluginInstall. With submodules, one would have to go into the cloned repo and run submodule init and update commands by hand.
alphaCTzo7G commentedon Nov 9, 2017
I have been using python-mode, with the default configuration for a couple of weeks. Thanks for creating this.
Couple of suggestions which may make it more popular and lead to continued development and improvements:
Why don't we advertise it as such, such as "Python IDE for VIM"?
The number of contributors to a project like this will be proportional to the number of users. With more user-friendly UIs like Pycharm, most new Python users are going to Pycharm. I wonder why we have such a lengthy first page. This would lead to the "paradox of choice" https://en.wikipedia.org/wiki/The_Paradox_of_Choice. Here's the launch page for Tensorflow https://github.com/tensorflow/tensorflow. I am pretty sure Tensorflow is way more complicated than python-mode. Maybe we can list the high-level functionalities provided by Pymode with some videos/demonstrations, and leave the rest to the wiki?
Neovim has moved to Apache 2.0 license. I have heard that a lot of companies prevent/discourage use of GPL licensed products, because of IP risk. http://corporate.findlaw.com/business-operations/the-risks-of-open-source-software.html. This is probably one of the reasons Neovim is changing their license. I would guess most programmers work in big corporations, and being GPL licensed, we automatically lose a big chunk of users. https://github.com/neovim/neovim/blob/master/README.md#license. Can we move to Apache 2.0/MIT license?
I notice that there are multiple libraries being used, some of which maybe redundant, not developed anymore or maybe there are better options available since 2015. http://liuchengxu.org/posts/use-vim-as-a-python-ide/. Can we use some of the more mordern ones?
In general, I have seen that VIM has tonnes of plugins for the same function, and it takes a while to figure out which one maybe best suited. Where python-mode could really add value for Python coders in VIM, is to screen all these plugins and provide, the best ones, great defaults, such that the user can get a IDE like environment quickly and get back to coding.
Thanks a lot for creating pymode. I can help with some of these.. or other ideas from the community over time.
fmv1992 commentedon Nov 9, 2017
@MadWombat,
tl;dr: submodules structure benefits outweighs its costs.
Those shortcomings are known but I think the benefits outweighs the costs in
this case. To update a single library usually means removing one folder and
replacing with another one full of changes that are tracked (and shouldn't be).
And this a mice and rat game because we will keep having to "paste" libraries
into our folder.
Another point is that in the makefile we are installing those libraries via pip
which I also think is not the best approach. Vim should not interfere with the
user's libraries.
And to remedy this issue one command is enough:
git submodule update --init --recursive && git submodule update --recursive --remoteMy idea is to move the tip of this branch to
master(thus keeping the currentlayout) and change
developto this new structure and keep it this way for 1-3months to improve user testing and collect feedback. Also the libs folder is
overcomplicated now and I suspect we don't use a good part of the code that is
there.
fmv1992 commentedon Nov 9, 2017
@alphaCTzo7G,
Those are very good points!
Brief comment:
Totally agree on this improvement of the plugin. Good point.
(Same argument as 1.) Redesign of the
readme.rstis needed.This point is hairy I think for two reasons.
First I don't think we can change the license without the author's consent
(I'm not the author).
Second I don't think this software clashes with corporate interests/IP
because you are only required to release source code if you distribute this
software as part of your product. In other words, corporations can use this
software to develop proprietary python applications and distribute them with
whatever licenses. If they distribute their product with embedded
python-mode in it then they have to distribute the source code too (which
I hardly think is the case).
I will put this in our roadmap. The libs folder is still obscure to me (in
terms of what is actually used and what is not) and mapping them is
a priority.
Great feedback! And I also thank those who created python mode :)
songww commentedon Nov 9, 2017
python-mode is a great. Thanks for your work.
There are many plugins for flake8, but submodule or sub-folder is too hard to use that.
When I register flake8 to git hooks, it will has two copies(maybe version also is different).
Use the user's library as an option?
English is not my native laguage.
fmv1992 commentedon Nov 9, 2017
@songww,
If I understood correctly you also oppose the use of submodules.
Again I think that the benefits outweigh the costs as a single command has to
be issued and the user can either decide which version of each library he wants
to use.
Also I don't want vim to interfere with a users choice of python packages (pip
packages).
Flake8 sounds like a good idea. But before we start adding more
packages/complexity to our mix I think we should focus on what we have right
now.
Thus I suggest that we create a list of linters/tools and discuss on those that
should be implemented. But again our starting point should be what we have
right now and have them operating correctly.
Good feedback!
fmv1992 commentedon Nov 14, 2017
The most pressing matter is rewriting the
readmefile I guess.I'll give it a try on 2017/11/15 ; by no means wait for me to improve it!
alphaCTzo7G commentedon Nov 14, 2017
I will work on the readme a bit over the next few days.. I have been going over the options for upgrading the libraries over the weekend a bit:
There have been a lot of interesting developments going on in the development of "Language server protocols", which might unite the need for separate libraries for separate editors and languages.. but I am not sure how far out that is: https://github.com/prabirshrestha/vim-lsp
It seems that almost universally people are using jedi instead of rope for autocompletion
In terms of linting docstrings, pep257 has be renamed/upgraded to pydocstyle.
There's ide's based on Neovim;https://github.com/onivim/oni
for formatting, Google/yapf seems to be more popular than autopep8.
a lot of the complaints regarding using pymode vs IDEs is that the lack of async in VIM 7.4 makes it slow to manage large libraries. So people switch to Pycharm.
For refactoring, rope is pretty much all there is for VIM.
However, none of these are well integrated like python-mode. You have to do all of it on a one by one basis..
fmv1992 commentedon Nov 14, 2017
@alphaCTzo7G,
I'll go through each point later but something occurred to me.
Assume the following are true:
Then I think that the following is a conclusion: each user will setup its own
code completion tool.
Thus I think that we can in the future add a robust completion feature that
should be off by default (let's go with jedi if what you said is true) but for
now assume that the users will have their own completions set up.
For instance I have tried both NeoComplete
https://github.com/Shougo/neocomplete.vim and YouCompleteMe
https://valloric.github.io/YouCompleteMe/ and approved both.
The idea behind 'outsourcing' this is to let us focus on the most pressing
needs that we have right now (the
readme.md, the roadmap for neededmaintenances and then adding new features IMO; adding support for jedi being
part of the third group).
See a related issue here:
#806
updated readme.rst to readme.md according to #805 and discussions
changed readme.rst to readme.md according to issues #805 and #813
changed readme.rst to readme.md according to issues #805 and #813
alphaCTzo7G commentedon Nov 16, 2017
I agree we should add tests.
We can start with something simple like you had.
Havent use the following frameworks myself, but heard of them while looking for something else:
alphaCTzo7G commentedon Nov 16, 2017
In terms of modularity, we should probably try to delegate as much of the work as possible to third party plugins. That way we can reduce our work, and as and when those plugins improve, we can just pull the improvements into this one.
Before we decide for how to modify this library, so that its more useful for everybody and so that we can sustain it long term, should we come up with a list of features/specs that are necessary for a IDE-type workflow for a moderately sized project in VIM?
Perhaps, we can chose a reasonably large python project OS project (sklearn, tensorflow etc), and try to figure out whether the current plugin can meet the specs.
If not, what are the bottlenecks, and how to go about fixing that. Otherwise, we might be fixing problems that dont need to be fixed or dont add value.
What do you guys think?
fmv1992 commentedon Nov 18, 2017
I think so as well. I'm not sure the best way to track those needs though.
fmv1992 commentedon Nov 18, 2017
I have reworked the mechanism and output of troubleshooting (updated the readme.md and the documentation as well).
This will possibly make it even more helpful to diagnose problems.
fmv1992 commentedon Nov 19, 2017
There have been a lot of interesting developments going on in the development of "Language server protocols", which might unite the need for separate libraries for separate editors and languages.. but I am not sure how far out that is: https://github.com/prabirshrestha/vim-lsp
It seems that almost universally people are using jedi instead of rope for autocompletion
In terms of linting docstrings, pep257 has be renamed/upgraded to pydocstyle.
dev_submodulesmerging.There's ide's based on Neovim;https://github.com/onivim/oni
experience. However I'm not sure what the trade-offs are and I have not
tried neovim (nor oni).
And well, this is Vim, so when faced with any trade-offs between 'bells
and whistles' and vim's 'culture' so to speak we should go the vim-way.
for formatting, Google/yapf seems to be more popular than autopep8.
post!
a lot of the complaints regarding using pymode vs IDEs is that the lack of
async in VIM 7.4 makes it slow to manage large libraries. So people switch
to Pycharm.
a matter of 'better vs worse'. It is just a matter of style. It sure does
not mean that we rejoice in having things barebones-vim-like. We should
go after improvements and functionality but we will never be a GUI-IDE.
Let's keep this in mind.
For refactoring, rope is pretty much all there is for VIM.
Two weeks have passed since the opening of this post. I think we are able to
go with submodules, freeze this discussion for a while and work on this small
roadmap that we have agreed upon.
What do you think?
fmv1992 commentedon Nov 20, 2017
Dear python-mode community,
I'm super glad to announce that we now have a test suite up and running!
More important than having the Travis CI badge:
Is to impart reliability to our software by creating a meaningful test
suite.
We now have tests that despite its complicatedness test python-mode's most
important functions.
The documentation also got a new section on Development.
None of those additions are set in stone but they represent a milestone in our
projects governance.
Best,
diraol commentedon Nov 23, 2017
Hi folks,
A little bit late on the discussion (as I've told @fmv1992, I was traveling on the last weeks)...
What about using git subtree instead of git submodules?
http://endot.org/2011/05/18/git-submodules-vs-subtrees-for-vim-plugins/
I haven't used it my self, but it looks like a nice way to manage 'external dependencies'.
This would add some extra effort to the development community (but not the craziness of copying files around), make the git history a little bit dirtier, but would ease the life of our users, avoiding the need of using extra git commands to have the plugin working.
Does it make any sense?
fmv1992 commentedon Dec 1, 2017
@diraol,
I cannot access this website right not but from I read on:
I think submodules is what we are looking for. We already have 6 or 7 plugins
and this number is likely to increase in the future. As we have seen those
changes from one version to another can be several hundreds of lines long.
The conclusion that I have is that if I want to track pymode's file I'll have
a lot of difficulty.
One command can sometimes be a hindrance but I think it still wins over
subtrees.
We could try that on a different branch in the future and see how it goes :)
stale commentedon Apr 30, 2018
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
stale commentedon May 21, 2018
Closed due to inactivity.