pydev-code

[Pydev-code] Some ideas about PyDev

[Radim K. <ra...@ku...>]

Hi Fabio and all,

I have a few ideas what I would like to change improve in current
PyDev. Basically most of them are a result of my attempts to improve
Python delelopment experience in our environment. I really like the
fact that there is already so much work done. Thanks for that. Still
there is some space for improvements. Some of them are general and
some are more specific for my use cases.

- speed and support for included/excluded resources: currently the
initial scan of platform and project takes too long + it tries to scan
recursively everything under the path root directories. This is very
unfortunate if your project is using SCM shared by more projects and
you really do not want to analyze unrelated sources.

- coverage: currently there are several launch types (run, (unit)test,
coverage) and two modes (exec/debug) while more natural to me is two
launch types (run / test) and three modes (exec/debug/coverage). Note
that this is similar to an approach taken by popular plugin for Java
coverage EclEmma (http://eclemma.org/). Also the UI provided by
current coverage supprt can by improved a lot - I'd like to see the
result as some kind of annotation in editor

- follow more Eclipse API/SPIs. Now I have some troubles as PyDev
converts eclipse Path to java.io.File almost instantly. This makes it
hard to use Eclipse links to link only to a specific directories (I do
that to work around a problem mentioned above).

I can submit patches for some of these as well as do some bugfixes if
there is an agreement that these are good ideas. Comments, feedback?

I'd also like to know what is current estimation for Ganymede (Eclipse
3.4) support?

-Radim

Re: [Pydev-code] Some ideas about PyDev

[Fabio Z. <fa...@gm...>]

Hi Radim,

Getting back to answer some e-mails (as the 1.3.19 is finally out)...

- speed and support for included/excluded resources: currently the
> initial scan of platform and project takes too long + it tries to scan
> recursively everything under the path root directories. This is very
> unfortunate if your project is using SCM shared by more projects and
> you really do not want to analyze unrelated sources.


I think that the initial scan should be much faster now (with the
improvements on pydev extensions)... also, I think it's a bit strange on
having the source that way if they should be unrelated (can you give some
examples on the organization and why it's like that -- so that I can
understand better what's really needed for solving what you want).

- coverage: currently there are several launch types (run, (unit)test,
> coverage) and two modes (exec/debug) while more natural to me is two
> launch types (run / test) and three modes (exec/debug/coverage). Note
> that this is similar to an approach taken by popular plugin for Java
> coverage EclEmma (http://eclemma.org/). Also the UI provided by
> current coverage supprt can by improved a lot - I'd like to see the
> result as some kind of annotation in editor


Improvements in the coverage are surely welcome!.

- follow more Eclipse API/SPIs. Now I have some troubles as PyDev
> converts eclipse Path to java.io.File almost instantly. This makes it
> hard to use Eclipse links to link only to a specific directories (I do
> that to work around a problem mentioned above).


Yes, that's a major refactoring to be done... (that's why currently the
links are only actually followed for the source directories)

I can submit patches for some of these as well as do some bugfixes if
> there is an agreement that these are good ideas. Comments, feedback?


For the java.io.File fix, I'm not very sure (I'd probably like to do it
myself, as it's probably going to work on lots of places in the code -- and
unit-tests), as for the others, I'm all for it.

I'd also like to know what is current estimation for Ganymede (Eclipse
> 3.4) support?
>

It's actually mostly supported, but official support is waiting for
https://sourceforge.net/tracker/index.php?func=detail&aid=2002870&group_id=85796&atid=577329to
be fixed (a patch for that one would be pretty nice).

Cheers,

Fabio

Re: [Pydev-code] Some ideas about PyDev

[Radim K. <ra...@ku...>]

Comments below.

On Fri, Aug 1, 2008 at 12:07 PM, Fabio Zadrozny <fa...@gm...> wrote:
> Hi Radim,
>
> Getting back to answer some e-mails (as the 1.3.19 is finally out)...
>
>> - speed and support for included/excluded resources: currently the
>> initial scan of platform and project takes too long + it tries to scan
>> recursively everything under the path root directories. This is very
>> unfortunate if your project is using SCM shared by more projects and
>> you really do not want to analyze unrelated sources.
>
> I think that the initial scan should be much faster now (with the
> improvements on pydev extensions)... also, I think it's a bit strange on
> having the source that way if they should be unrelated (can you give some
> examples on the organization and why it's like that -- so that I can
> understand better what's really needed for solving what you want).
>
Few ideas of an arrangement(s) that can lead you to that situation:

- you have more project that people work on (+ various internally
developed libraries)

- there is one large repository holding all sources (to encourage reuse)

- there are not only Python source but also other languages mixed but
for some (usually historical) reason root of python path is also a
parent of folders where Java/C/C++/... sources are stored as well

- some of those above + to avoid large checkout/view of repository you
can use a view over the whole repository

And yes, analysis phase seems to be faster now.

-Radim

>
> Cheers,
>
> Fabio
>