Add python script that generates gmock_mutant.h...

Add python script that generates gmock_mutant.h
Resolve ambiguity - error C2782 in VisualC or "no matching function to call" in gcc - when attempt to create functor for base class member function.
Support for free (non class members) functions.

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=31848

[amit, 2009-11-07 05:17:41]

for gmock_mutant.h we also need __stdcall versions to be generated and don't
forget to wrap them in #ifdef OS_WIN :)

[awong, 2009-11-09 21:57:21]

Looks reasonable.  But I'd like to see the template pulled out into constants so
that they don't muddy up the code quite as much.

Also, for the Expand function, why aren't we just passing in a string that you
can use the % operator on?

http://codereview.chromium.org/371046/diff/3002/3003
File testing/generate_gmock_mutant.py (right):

http://codereview.chromium.org/371046/diff/3002/3003#newcode1
Line 1: #!/usr/bin/python
Eek...can't have a space before the #. That'll break things.

http://codereview.chromium.org/371046/diff/3002/3003#newcode8
Line 8: 
remove extra newline please.

http://codereview.chromium.org/371046/diff/3002/3003#newcode13
Line 13: n = 1 + line[:width].rfind(",")
What if there is no comma within |width| characters?

http://codereview.chromium.org/371046/diff/3002/3003#newcode14
Line 14: # Assume there is a space after the comma
Periods at end of sentences.  Here and in the rest of the doc strings. :)

http://codereview.chromium.org/371046/diff/3002/3003#newcode31
Line 31: """Cleans artifacts from generated C++ code."""
This needs a bit more explanation for why these changes have to happen.

http://codereview.chromium.org/371046/diff/3002/3003#newcode43
Line 43: def Gen(pattern, n):
Gen, GenAlpha are a bit non-descript.  Make them a bit more descriptive? Or add
a comment?

http://codereview.chromium.org/371046/diff/3002/3003#newcode170
Line 170: namespace testing {"""
For these blocks of inlined C code, move them into a set of constants at the top
of the file.  Easier to read.

http://codereview.chromium.org/371046/diff/3002/3003#newcode303
Line 303: print "inline R DispatchToMethod(T* obj, Method method,"
Can we use a big template like:
DISPATH_TO_METHOD = """inline R DispatchToMethod(T* obj, Method moethod,
                             const %(preboud)s p,
                             const %(calltime)s c) {
  return (obj->*method)(%(args)s);
}
"""

Then you can print with something like:
print DISPATH_TO_TEMPLATE % {
   'prebuild': GenTuple("P%", prebound),
   'calltime': GenTuple("C%", calltime),
   'args': Merge([GenAlpha("p.%", prebound), GenAlpha("c.%", calltime)])
}

http://codereview.chromium.org/371046/diff/3002/3003#newcode377
Line 377: print "#endif  // GMOCK_MUTANT_INCLUDE_LATE_OBJECT_BINDING\n"
I think python style guide says 2 newlines between top-level declarations.

http://codereview.chromium.org/371046/diff/3002/3003#newcode381
Line 381: for prebound in range(0, 4 + 1):
xrange is prefered for loop iterators.

[awong, 2009-11-09 21:58:48]

http://codereview.chromium.org/371046/diff/4002/4003
File testing/generate_gmock_mutant.py (right):

http://codereview.chromium.org/371046/diff/4002/4003#newcode40
Line 40: return [pattern.replace("%", x) for x in it]
Yeah..here....why aren't we just saying hte pattern string should be in the
form:

"%s whatever" and just doing a [ pattern % x for x in it ]?

[stoyan, 2009-11-10 21:30:23]

On Mon, Nov 9, 2009 at 1:58 PM,  <ajwong@chromium.org> wrote:
> http://codereview.chromium.org/371046/diff/4002/4003#newcode40
> Line 40: return [pattern.replace("%", x) for x in it]
> Yeah..here....why aren't we just saying hte pattern string should be in
> the form:
>
> "%s whatever" and just doing a [ pattern % x for x in it ]?

I think usage of Gen("A%", n) instead of "Gen("A%s", n) is somewhat
clearer in the light of the humble intentions of the pattern
expansion.

Uploaded a new version. Hope looks better. Added __stdcall methods as
Amit requested.

[stoyan, 2009-11-11 23:32:33]

ping?

On Tue, Nov 10, 2009 at 1:29 PM, stoyan <stoyan@chromium.org> wrote:
> On Mon, Nov 9, 2009 at 1:58 PM, =A0<ajwong@chromium.org> wrote:
>> http://codereview.chromium.org/371046/diff/4002/4003#newcode40
>> Line 40: return [pattern.replace("%", x) for x in it]
>> Yeah..here....why aren't we just saying hte pattern string should be in
>> the form:
>>
>> "%s whatever" and just doing a [ pattern % x for x in it ]?
>
> I think usage of Gen("A%", n) instead of "Gen("A%s", n) is somewhat
> clearer in the light of the humble intentions of the pattern
> expansion.
>
> Uploaded a new version. Hope looks better. Added __stdcall methods as
> Amit requested.
>

[amit, 2009-11-11 23:37:39]

lgtm

[awong, 2009-11-12 00:28:08]

Looks a lot nicer.  It'd be nice to see the rest of the multiline strings pulled
out though.

http://codereview.chromium.org/371046/diff/7003/8003
File testing/generate_gmock_mutant.py (right):

http://codereview.chromium.org/371046/diff/7003/8003#newcode6
Line 6: import string
newline after the import.

http://codereview.chromium.org/371046/diff/7003/8003#newcode313
Line 313: def PrintHeader():
Just inline the code for these functions?  They're a bit too trivial.

http://codereview.chromium.org/371046/diff/7003/8003#newcode374
Line 374: method_functor = """template <typename R, typename T, typename U,
%(tp)s>
I'd still prefer to see these pulled out into constants, with a comment
explaining what parameters are expected from the hash.

Basically, anything that messes with the indent level would be good to pull out
into a comment.

[stoyan, 2009-11-12 18:51:29]

http://codereview.chromium.org/371046/diff/7003/8003
File testing/generate_gmock_mutant.py (right):

http://codereview.chromium.org/371046/diff/7003/8003#newcode6
Line 6: import string
On 2009/11/12 00:28:08, awong wrote:
> newline after the import.

Done.

http://codereview.chromium.org/371046/diff/7003/8003#newcode313
Line 313: def PrintHeader():
On 2009/11/12 00:28:08, awong wrote:
> Just inline the code for these functions?  They're a bit too trivial.

Done.

http://codereview.chromium.org/371046/diff/7003/8003#newcode374
Line 374: method_functor = """template <typename R, typename T, typename U,
%(tp)s>
On 2009/11/12 00:28:08, awong wrote:
> I'd still prefer to see these pulled out into constants, with a comment
> explaining what parameters are expected from the hash.
> 
> Basically, anything that messes with the indent level would be good to pull
out
> into a comment.

Done.

[awong, 2009-11-12 23:08:25]

LGTM

Thanks for making the changes!