Weird behaviour building Pygr epydocs with Fink

1 view
Skip to first unread message

Marek Szuba

unread,
Dec 10, 2009, 9:53:33 PM12/10/09
to pygr...@googlegroups.com
Hello there,

https://sourceforge.net/tracker/index.php?func=detail&aid=2859512&group_id=17203&atid=414256

Please see monipol's comment from the 9th of December. As I responded
to him most errors produced during doc building are harmless, it is
however somewhat surprising to see that installing a certain version of
Pygr causes epydoc to begin behaving differently... In case you
haven't had a chance to compare the two log files, what exactly happens
is that with pygr-py26 installed, during processing of the apps/
subdirectory epydoc begins believing SeqPath and TupleO come from seqdb
instead of sequence and sqlgraph, respectively.

Have you guys seen anything like this before or suspect where this
behaviour could come from? Personally I suspect a problem with epydoc
and not Pygr but perhaps there is something subtle at work here.

Cheers,
--
MS

Istvan Albert

unread,
Dec 11, 2009, 11:06:10 AM12/11/09
to pygr-dev


On Dec 10, 9:53 pm, Marek Szuba <mare...@gmail.com> wrote:

> Have you guys seen anything like this before or suspect where this
> behaviour could come from? Personally I suspect a problem with epydoc
> and not Pygr but perhaps there is something subtle at work here.

Epydoc uses two ways to generate the documentation. One is parsing the
docstrings, the other it to import the module and use introspection to
identify what goes where. Python being a dynamic language one could be
adding functions/classes during runtime. Sometimes the two behaviors
can clash and produce some headscratching results. I had some problems
like that.

You can turn off introspection with a command line flag to epydoc,
that would probably do away with the errors that pop up.

best,

Istvan

Christopher Lee

unread,
Dec 11, 2009, 6:13:13 PM12/11/09
to pygr...@googlegroups.com

On Dec 10, 2009, at 6:53 PM, Marek Szuba wrote:

> epydoc begins believing SeqPath and TupleO come from seqdb
> instead of sequence and sqlgraph, respectively.

I think this is because seqdb imports * from those modules:

from sequence import * # @CTB
from sqlgraph import * # @CTB

We could limit what symbols get included in the seqdb module namespace as is standard practice for Python modules.

-- Chris
Reply all
Reply to author
Forward
0 new messages