From mscott at goldenspud.com  Thu Mar  3 20:39:21 2005
From: mscott at goldenspud.com (Matthew Scott)
Date: Thu, 03 Mar 2005 13:39:21 -0600
Subject: [py-dev] using Python profiler with py.test
Message-ID: <422767E9.7030105@goldenspud.com>


I'm having problems using the Python profiler with py.test

If I write a simple one-liner as 'hello.py':

   import py; print py.path.local()

Then run "python2.4 -m profile hello.py", I get a report from the profiler.

However, if I run "python2.4 -m profile bin/py.test" from within the py 
directory, I get the output below.

Any ideas or tips on using the Python profiler with py.test?


inserting into sys.path: /home/gldnspud/s/py-dist
============================= test process starts 
=============================
testing-mode: inprocess
executable  : /usr/bin/python2.4  (2.4.1-alpha-0)
using py lib: /home/gldnspud/s/py-dist/py <rev unknown>
Traceback (most recent call last):
   File "/usr/lib/python2.4/profile.py", line 611, in ?
     run('execfile(%r)' % (sys.argv[0],), options.outfile, options.sort)
   File "/usr/lib/python2.4/profile.py", line 72, in run
     prof = prof.run(statement)
   File "/usr/lib/python2.4/profile.py", line 448, in run
     return self.runctx(cmd, dict, dict)
   File "/usr/lib/python2.4/profile.py", line 454, in runctx
     exec cmd in globals, locals
   File "<string>", line 1, in ?
   File "bin/py.test", line 5, in ?
     main()
   File "/home/gldnspud/s/py-dist/py/test/cmdline.py", line 32, in main
     run.inprocess(args, filenames)
   File "/home/gldnspud/s/py-dist/py/test/run.py", line 141, in inprocess
     driver.run(fncollectors)
   File "/home/gldnspud/s/py-dist/py/test/drive.py", line 41, in run
     self.reporter.start()
   File "/home/gldnspud/s/py-dist/py/test/report/text/reporter.py", line 
53, in start
     self.out.line("initial testconfig %d: %s" %(i, x))
   File "/home/gldnspud/s/py-dist/py/path/extpy/extpy.py", line 38, in 
__str__
     def __str__(self):
   File "/usr/lib/python2.4/profile.py", line 238, in trace_dispatch_i
     if self.dispatch[event](self, frame, t):
   File "/usr/lib/python2.4/profile.py", line 295, in trace_dispatch_call
     assert (self.cur is None or \
AssertionError: ('Bad call', ('/home/gldnspud/s/py-dist/py/initpkg.py', 
149, '__getattr__'))


-- 
Matthew R. Scott


From hpk at trillke.net  Fri Mar  4 08:06:15 2005
From: hpk at trillke.net (holger krekel)
Date: Fri, 4 Mar 2005 08:06:15 +0100
Subject: [py-dev] using Python profiler with py.test
In-Reply-To: <422767E9.7030105@goldenspud.com>
References: <422767E9.7030105@goldenspud.com>
Message-ID: <20050304070615.GG3663@solar.trillke.net>

Hi Matthew! 

On Thu, Mar 03, 2005 at 13:39 -0600, Matthew Scott wrote:
> I'm having problems using the Python profiler with py.test
> 
> If I write a simple one-liner as 'hello.py':
> 
>   import py; print py.path.local()
> 
> Then run "python2.4 -m profile hello.py", I get a report from the profiler.
> 
> However, if I run "python2.4 -m profile bin/py.test" from within the py 
> directory, I get the output below.
> 
> Any ideas or tips on using the Python profiler with py.test?

Hum, i can reproduce your problem but am not sure exactly what is
going on.  If i take out the triggering assertion in profile.py 
then it seems to work.  Maybe the assertion somehow triggers 
because of the lazy module loading in py/initpkg.py.  Forgive me, 
but I am not in the mood currently to examine how profile.py works ... 

Anyway, i think that we should rather try to incorporate 
hotshot rather than profile directly into py.test as this is the 
recommended way of doing profiling these days ASFAIK.  (Btw, I assume 
we want to profile the tested application and not py.test itself ...)
Do you have a suggestion for a cmdline-API? 

cheers, 

    holger


From mscott at goldenspud.com  Fri Mar  4 12:30:48 2005
From: mscott at goldenspud.com (Matthew Scott)
Date: Fri, 04 Mar 2005 05:30:48 -0600
Subject: [py-dev] using Python profiler with py.test
In-Reply-To: <20050304070615.GG3663@solar.trillke.net>
References: <422767E9.7030105@goldenspud.com>
	<20050304070615.GG3663@solar.trillke.net>
Message-ID: <422846E8.3090505@goldenspud.com>

holger krekel wrote:
> Anyway, i think that we should rather try to incorporate 
> hotshot rather than profile directly into py.test as this is the 
> recommended way of doing profiling these days ASFAIK.  (Btw, I assume 
> we want to profile the tested application and not py.test itself ...)
> Do you have a suggestion for a cmdline-API? 

D'oh!  I haven't tried hotshot yet.  This was my first attempt at 
profiling something in Python so I just gravitated toward something that 
had was based on the same root word :)

I'll check out hotshot, since I haven't yet... :)


-- 
Matthew R. Scott


From dialtone at divmod.com  Sun Mar  6 01:44:51 2005
From: dialtone at divmod.com (Valentino Volonghi aka Dialtone)
Date: Sun, 06 Mar 2005 01:44:51 +0100
Subject: [py-dev] py.test patches and problems
Message-ID: <422A5283.60906@divmod.com>

I think there is far too magic in py.test :).

Since I need a markdown parser I've started a new project. Unfortunately 
the Markdown test suite is completely useless (whoever wrote it must 
have some problems with blank lines in html output).

Hence I wanted to build my own test suite using py.test

First problem: I'm under MacOSX and I'm using python built with the 
framework. This means _findpy.py is not ready to locate my py dir and 
fails (even if import py finishes without problems).
I modified _findpy.py in this way:

Index: py/bin/_findpy.py
===================================================================
--- py/bin/_findpy.py   (revision 9671)
+++ py/bin/_findpy.py   (working copy)
@@ -27,6 +27,9 @@
 if not searchpy(abspath(os.curdir)):
     if not searchpy(opd(abspath(sys.argv[0]))):
         if not searchpy(opd(__file__)):
-            raise SystemExit, "Could not find 'py' package!"
+            for path in sys.path:
+                if searchpy(path):
+                    break
+            else: raise SystemExit, "Could not find 'py' package!"
 
 import py


And now it works. Pythondir under MacOSX with framework is:
/Library/Frameworks/Python.framework/Versions/2.4/lib/python2.4/site-packages
while python executable is under:
/Library/Frameworks/Python.framework/Versions/2.4/bin/python

But then there is another problem with other software that uses 
__import__() function.

I'm working with Nevow and it uses a particular function: namedAny which 
loads a module from PYTHONPATH using the complete path to the module.

This is one of the many error outputs that I have while running 'py.test'.

    def namedAny(name):
        """Get a fully named package, module, module-global object, or 
attribute.
        """
        names = name.split('.')
        topLevelPackage = None
        moduleNames = names[:]
        while not topLevelPackage:
            try:
                trialname = '.'.join(moduleNames)
E               topLevelPackage = __import__(trialname)
~               ImportError: 
__import__('nevow.flat.flatstan.TagSerializer')  << No module named 
TagSerializer

Is there any way to fix this last behaviour? I fear that import hooks 
are too magic, for a py.test newbie like me, to dig in.

-- 
Valentino Volonghi aka Dialtone
Now Running MacOSX 10.3.8
Blog: http://vvolonghi.blogspot.com
http://weever.berlios.de



From hpk at trillke.net  Sun Mar  6 11:10:54 2005
From: hpk at trillke.net (holger krekel)
Date: Sun, 6 Mar 2005 11:10:54 +0100
Subject: [py-dev] py.test patches and problems
In-Reply-To: <422A5283.60906@divmod.com>
References: <422A5283.60906@divmod.com>
Message-ID: <20050306101054.GP3663@solar.trillke.net>

Hi Valentino, 

On Sun, Mar 06, 2005 at 01:44 +0100, Valentino Volonghi aka Dialtone wrote:
> I think there is far too magic in py.test :).

maybe.  We try to keep it down and only do neccessary magic ... 

> Since I need a markdown parser I've started a new project. Unfortunately 
> the Markdown test suite is completely useless (whoever wrote it must 
> have some problems with blank lines in html output).
> 
> Hence I wanted to build my own test suite using py.test
> 
> First problem: I'm under MacOSX and I'm using python built with the 
> framework. This means _findpy.py is not ready to locate my py dir and 
> fails (even if import py finishes without problems).

Hum, did you put "py.test" somewhere outside the py library? 
I am using MacOSX (both framework and non-framework installs), 
without any problems.  

> I modified _findpy.py in this way:
> 
> Index: py/bin/_findpy.py
> ===================================================================
> --- py/bin/_findpy.py   (revision 9671)
> +++ py/bin/_findpy.py   (working copy)
> @@ -27,6 +27,9 @@
> if not searchpy(abspath(os.curdir)):
>     if not searchpy(opd(abspath(sys.argv[0]))):
>         if not searchpy(opd(__file__)):
> -            raise SystemExit, "Could not find 'py' package!"
> +            for path in sys.path:
> +                if searchpy(path):
> +                    break
> +            else: raise SystemExit, "Could not find 'py' package!"

I just committed a simpler "fix" where last 'if' just
results in a "pass" so that "import py" will be executed
either way. 

> But then there is another problem with other software that uses 
> __import__() function.

strange. 

> I'm working with Nevow and it uses a particular function: namedAny which 
> loads a module from PYTHONPATH using the complete path to the module.
> 
> This is one of the many error outputs that I have while running 'py.test'.
> 
>    def namedAny(name):
>        """Get a fully named package, module, module-global object, or 
> attribute.
>        """
>        names = name.split('.')
>        topLevelPackage = None
>        moduleNames = names[:]
>        while not topLevelPackage:
>            try:
>                trialname = '.'.join(moduleNames)
> E               topLevelPackage = __import__(trialname)
> ~               ImportError: 
> __import__('nevow.flat.flatstan.TagSerializer')  << No module named 
> TagSerializer
> 
> Is there any way to fix this last behaviour? I fear that import hooks 
> are too magic, for a py.test newbie like me, to dig in.

I presume there must be a much larger traceback. I just looked at 
the nevow import logic and have the impression that it's  
at least as magic as the py lib's :-) 

Anyway, can you try to replace line 337 of py/path/common.py from 

    mod.__file__ = PathStr(self)
to 
    mod.__file__ = str(self)  

and see if that fixes your problem?  

thanks, 

    holger


From jan at balster.info  Mon Mar 14 15:48:59 2005
From: jan at balster.info (Jan Balster)
Date: Mon, 14 Mar 2005 15:48:59 +0100
Subject: [py-dev] Bug in collect.Error
Message-ID: <4235A45B.5070906@balster.info>

Hello,

this is my first post to py-dev.
Thanks for py.test!

I found an Bug in collect.Error.__repr__[line 27]:

calling __repr__ raises an exception:

     def __repr__(self):
E       tb = list(self.excinfo)[-1]
~       TypeError:   << iteration over non-sequence


__repr__ should be changed to:

def __repr__(self):
        tb = list(self.excinfo.traceback)[-1]
                               ^^^^^^^^^
        return '<collect.Error in %r, line %d>' % (tb.frame.code.path,
                                                   tb.lineno+1)

Testcase:

import py
def test_Error_repr_returns_string():
    try:
        raise TypeError
    except:
        import sys
        error = py.test.collect.Error(sys.exc_info())
    import types
    assert isinstance(repr(error), types.StringTypes)



Jan Balster




From edvard.majakari at staselog.com  Tue Mar 15 14:07:25 2005
From: edvard.majakari at staselog.com (Edvard Majakari)
Date: Tue, 15 Mar 2005 15:07:25 +0200
Subject: [py-dev] Creating test skeleton automatically from modules
Message-ID: <87zmx5njle.fsf@titan.staselog.com>


Hi,

I wrote a simple program which uses Python's compiler module to parse given
set of modules and writes a py.test unit test skeleton for each
module. Currently it is not too clever, mostly because I don't have very clear
ideas how unit testing should be done. However, it does create code for
automatically generating test methods :)

Say, if you have module bar and therein method named foo, pytestgen will
automatically create 

TestBar:

  ...

  def test_foo(self):

      known_values(,)

      for arg1, arg2, expected in known_values:
          yield self.foo, arg1, arg2, expected

  def foo(self, arg1, arg2, expectd):
 
      assert bar.foo(arg1, arg2) == expected


I intend to develop the program further, hopefully integrate it to Eclipse
pydev plugin :)

If you're interested in trying it out, see

http://majakari.net/dl/pytestgen-0.11.tar.bz2

It's not too advanced (yet), but I already use it myself for my own purposes.

-- 
# Edvard Majakari		Software Engineer
# PGP PUBLIC KEY available    	Soli Deo Gloria!

$_ = '456476617264204d616a616b6172692c20612043687269737469616e20'; print
join('',map{chr hex}(split/(\w{2})/)),uc substr(crypt(60281449,'es'),2,4),"\n";



From hpk at trillke.net  Tue Mar 15 14:21:45 2005
From: hpk at trillke.net (holger krekel)
Date: Tue, 15 Mar 2005 14:21:45 +0100
Subject: [py-dev] Bug in collect.Error
In-Reply-To: <4235A45B.5070906@balster.info>
References: <4235A45B.5070906@balster.info>
Message-ID: <20050315132145.GI17456@solar.trillke.net>

Hi Jan, 

On Mon, Mar 14, 2005 at 15:48 +0100, Jan Balster wrote:
> Hello,
> 
> this is my first post to py-dev.
> Thanks for py.test!
> 
> I found an Bug in collect.Error.__repr__[line 27]:

thanks for pointing it out along with the test! 

i just committed this to the trunk although this will
likely become obsolete soon when the "py-collect" branch is 
merged back into the trunk (intended to unify and clean up
the test collection process) ... 

cheers, 

    holger


From jan at balster.info  Tue Mar 15 20:02:07 2005
From: jan at balster.info (Jan Balster)
Date: Tue, 15 Mar 2005 20:02:07 +0100
Subject: [py-dev] Bug in collect.Error
In-Reply-To: <20050315132145.GI17456@solar.trillke.net>
References: <4235A45B.5070906@balster.info>
	<20050315132145.GI17456@solar.trillke.net>
Message-ID: <4237312F.3070906@balster.info>

hi Holger,

holger krekel wrote:

> i just committed this to the trunk although this will
> likely become obsolete soon when the "py-collect" branch is 
> merged back into the trunk (intended to unify and clean up
> the test collection process) ... 

i just finished a small tkinter gui for py.test with a tree view for all
collectors and items.
the main part is a gui-reporter, which sends information through a
extpy.channel to the gui. i think, i have to use a subclass of driver now.

the new FailureCollector in run.py is great. i had a problem to get the
right name (like collector.listnames()) for a failed item returned by
the old FailureCollector. thats much easier now, thank you.

what kind of object is root in FailureCollector:
   "for root, names in failures"?
i think, it should be a collect.Directory or collect.Module, but sending
them through a extpy.channel doesn't work:
eval(repr(directory_object)) != directory_object

terminaldriver.run() (--> driver.run()) returns nothing (None), so i
can't deduce the object type from the source.

cheers,

    jan


From hpk at trillke.net  Tue Mar 15 21:29:07 2005
From: hpk at trillke.net (holger krekel)
Date: Tue, 15 Mar 2005 21:29:07 +0100
Subject: [py-dev] Bug in collect.Error
In-Reply-To: <4237312F.3070906@balster.info>
References: <4235A45B.5070906@balster.info>
	<20050315132145.GI17456@solar.trillke.net>
	<4237312F.3070906@balster.info>
Message-ID: <20050315202907.GA30690@solar.trillke.net>

Hi Jan, 

On Tue, Mar 15, 2005 at 20:02 +0100, Jan Balster wrote:
> holger krekel wrote:
> > i just committed this to the trunk although this will
> > likely become obsolete soon when the "py-collect" branch is 
> > merged back into the trunk (intended to unify and clean up
> > the test collection process) ... 
> 
> i just finished a small tkinter gui for py.test with a tree view for all
> collectors and items.

Cool! 

> the main part is a gui-reporter, which sends information through a
> extpy.channel to the gui. i think, i have to use a subclass of driver now.
> 
> the new FailureCollector in run.py is great. i had a problem to get the
> right name (like collector.listnames()) for a failed item returned by
> the old FailureCollector. thats much easier now, thank you.

hey, you are using the collect branch already?  Nice! and crazy :-) 

> what kind of object is root in FailureCollector:
>    "for root, names in failures"?

Well, this particular bit is not fully spelled out yet, 
to be honest. 

> i think, it should be a collect.Directory or collect.Module, but sending
> them through a extpy.channel doesn't work:
>
> eval(repr(directory_object)) != directory_object

yip, you are right.  Probably it will be "implicitely" encoded 
so that the FrontendDriver and the remote end will agree that
"root" is e.g. a filesystem path (as a string) which has to be 
wrapped by (indeed) a collect.Directory object. 
So the channel would only see lists and strings basically 
from which the "collect/item tree" can be constructed. 

> terminaldriver.run() (--> driver.run()) returns nothing (None), so i
> can't deduce the object type from the source.

Yes, that return value is one of the missing bits in the branch! 
I'll try to finish that up asap. 

cheers! 

    holger


From jan at balster.info  Wed Mar 16 15:59:19 2005
From: jan at balster.info (Jan Balster)
Date: Wed, 16 Mar 2005 15:59:19 +0100
Subject: [py-dev] branch py-collect was: Bug in collect.Error
Message-ID: <423849C7.4070104@balster.info>

hi holger,


> Yes, that return value is one of the missing bits in the branch! 
> I'll try to finish that up asap. 
> 

may i help?

below a diff for run.py and drive.py.
FrontendDriver(run.py) and Driver(drive.py) communicate only with lists
and tuples of strings.

hope this helps.

Jan


output of "svn diff run.py":
Index: run.py
===================================================================
--- run.py      (Revision 9787)
+++ run.py      (Arbeitskopie)
@@ -94,7 +94,7 @@
     args, failures = channel.receive()
     driver, paths = py.test.config.init(args, ignoreremote=True)
     if failures:
-        cols = [FailureCollector(failures)]
+        cols = FailureCollector(failures)
     else:
         cols = list(getcollectors(paths))
     driver.shouldclose = channel.isclosed
@@ -105,8 +105,15 @@
     def __init__(self, failures):
         self.failures = failures
     def __iter__(self):
-        for root, names in self.failures:
-            current = root
+        for rootpath, names in self.failures:
+            root = py.path.local(rootpath)
+            if root.check(dir=1):
+                current = py.test.Directory(root).Directory(root)
+            elif root.check(file=1):
+                current = py.test.Module(root).Module(root)
+            # root is fspath of names[0] -> pop names[0]
+            # slicing works with empty lists
+            names = names[1:]
             while names:
                 name = names.pop(0)
                 try:
@@ -143,8 +150,8 @@
                 break
             print "#" * 60
             print "# session mode: %d failures remaining" % len(failures)
-            for x in failures:
-                name = ":".join(x) # XXX
+            for root, names in failures:
+                name = ":".join(names) # XXX
                 print "Failure at: %r" % (name,)
             print "#    watching py files below %s" % rootdir
-            print "#                           ", "^" * len(rootdir)
+            print "#                           ", "^" * len(str(rootdir))

output of "svn diff drive.py":
Index: drive.py
===================================================================
--- drive.py    (Revision 9787)
+++ drive.py    (Arbeitskopie)
@@ -54,6 +54,11 @@
         except Exit, ex:
             pass

+        # return [(fspath as string, [names as string])]
+        return [(str(item.listchain()[0].fspath), item.listnames())
+                for item, res in self.getresults(py.test.Item.Failed)]
+
+
     def runone(self, colitem):
         if self.shouldclose():
             raise SystemExit, "received external close signal"



From hpk at trillke.net  Wed Mar 16 17:01:43 2005
From: hpk at trillke.net (holger krekel)
Date: Wed, 16 Mar 2005 17:01:43 +0100
Subject: [py-dev] branch py-collect was: Bug in collect.Error
In-Reply-To: <423849C7.4070104@balster.info>
References: <423849C7.4070104@balster.info>
Message-ID: <20050316160143.GH30690@solar.trillke.net>

Hi Jan, 

On Wed, Mar 16, 2005 at 15:59 +0100, Jan Balster wrote:
> hi holger,
> > Yes, that return value is one of the missing bits in the branch! 
> > I'll try to finish that up asap. 
> > 
> 
> may i help?

Sure :-)   

> below a diff for run.py and drive.py.
> FrontendDriver(run.py) and Driver(drive.py) communicate only with lists
> and tuples of strings.

applied, thanks!  

Btw, if you want to get an account with checkin rights just drop me
a private mail with your desired user name and public ssh key. 
I think that your perspective from a GUI side of things could help
to get py.test more mature and flexible! 

cheers, 

    holger


From jan at balster.info  Wed Mar 16 14:38:48 2005
From: jan at balster.info (Jan Balster)
Date: Wed, 16 Mar 2005 14:38:48 +0100
Subject: [py-dev] branch py-collect was: Bug in collect.Error
Message-ID: <423836E8.8000402@balster.info>

hi holger,


> Yes, that return value is one of the missing bits in the branch! 
> I'll try to finish that up asap. 
> 

may i help?

below a diff for run.py and drive.py.
FrontendDriver(run.py) and Driver(drive.py) communicate only with lists
and tuples of strings.

hope this helps.

Jan


output of "svn diff run.py":
Index: run.py
===================================================================
--- run.py      (Revision 9787)
+++ run.py      (Arbeitskopie)
@@ -94,7 +94,7 @@
     args, failures = channel.receive()
     driver, paths = py.test.config.init(args, ignoreremote=True)
     if failures:
-        cols = [FailureCollector(failures)]
+        cols = FailureCollector(failures)
     else:
         cols = list(getcollectors(paths))
     driver.shouldclose = channel.isclosed
@@ -105,8 +105,15 @@
     def __init__(self, failures):
         self.failures = failures
     def __iter__(self):
-        for root, names in self.failures:
-            current = root
+        for rootpath, names in self.failures:
+            root = py.path.local(rootpath)
+            if root.check(dir=1):
+                current = py.test.Directory(root).Directory(root)
+            elif root.check(file=1):
+                current = py.test.Module(root).Module(root)
+            # root is fspath of names[0] -> pop names[0]
+            # slicing works with empty lists
+            names = names[1:]
             while names:
                 name = names.pop(0)
                 try:
@@ -143,8 +150,8 @@
                 break
             print "#" * 60
             print "# session mode: %d failures remaining" % len(failures)
-            for x in failures:
-                name = ":".join(x) # XXX
+            for root, names in failures:
+                name = ":".join(names) # XXX
                 print "Failure at: %r" % (name,)
             print "#    watching py files below %s" % rootdir
-            print "#                           ", "^" * len(rootdir)
+            print "#                           ", "^" * len(str(rootdir))

output of "svn diff drive.py":
Index: drive.py
===================================================================
--- drive.py    (Revision 9787)
+++ drive.py    (Arbeitskopie)
@@ -54,6 +54,11 @@
         except Exit, ex:
             pass

+        # return [(fspath as string, [names as string])]
+        return [(str(item.listchain()[0].fspath), item.listnames())
+                for item, res in self.getresults(py.test.Item.Failed)]
+
+
     def runone(self, colitem):
         if self.shouldclose():
             raise SystemExit, "received external close signal"