Ignore changes in comments about tortoisesvn HOT 17 CLOSED

ExamDiff apparently has this feature implemented as a regex function so you can
define yourself comments for each programming language.

Or something like a SED in front of the diffing engine, i.e. have a stream 
editor like thingy alter the data before it's used in TMerge.
http://tortoisesvn.tigris.org/ds/viewMessage.do?dsForumId=4061&dsMessageId=29480
96

/*
this is my project
*/

Noticed that Viraj has claimed this; thanks.  This one's still on my wish list, 
mainly as the "SED in front of the diffing engine" option (Apr 14 2012 comment 
and link)  We do a lot of that stuff as manual pre-process steps to help 
compare large integration test baselines.  Our two most common pre-process use 
cases are:

1. Single line sed replacements to filter out things like GUIDs and timestamps
2. Full-file sorts to help distinguish sort order changes from other changes

Including use case #2 suggests making the pre-processor tool configurable, but 
that could be scope creep.  Just implementing #1 and someday adding a check-box 
to support #2 might yield be a better overall experience.

The hardest design questions here, in my mind, revolves around how custom 
filters are shared among users in an organization.  Right now our pre-processor 
scripts are largely a scattered collection of batch files, some checked into 
SVN (not necessarily known to anyone outside Team XYZ), others living on 
individual workstations, etc. One possible solution to this is problem would be 
to support a new  folder-level property, e.g. "tortoisemerge:customfilters" 
that pointed to a text (xml?) file (files?) that defines custom filters 
automatically available to anyone diffing a file in that folder or its 
subfolders.

this should not be implemented using external tools. Because let's face it: if 
you need them, you don't have them installed.

This should be done using regular expressions ( e.g., with regex_replace() ) to 
modify the files before they're fed to the diffing engine.

For the primary use case, I completely agree: 

1. choose a well-known, well-documented, regex syntax, support that, and
2. do it in a way that makes it easy for the guy down the hall to use the same 
quiver of regex filters that I use.

"Support for external filtering tools" is really an independent feature request 
(wrong forum I know; my apologies), but I mention it here since any future 
implementation of it would fit closely side-by-side with the current regex 
stuff.

By this, I mean support for pluggable external filters, not any actual new 
dependency on such filters.  For example, I just did a code review this morning 
of a freshly forked C# library where virtually all the changes were StyleCop 
driven formatting tweaks.  I (think I) could build a filtering tool using the 
Microsoft.CSharp namespace to convert a given C# file into a more canonical 
form to reduce a lot of this noise in a comparison.  Such a tool would probably 
work best with both input files as arguments, so one file could be treated as a 
reference source when resolving details like method re-names.

Thanks, Eric

I guess using a regex won't work with the current implementation since the 
whole file handling in TMerge is line based. And such regexes would need to 
work on multiple lines.

At least for now, I'll implement this using a simpler approach:
define line and block comments using tokens, for example for a c/c++ file that 
would be:

line comment start: //
block comment start: /*
block comment end: */

this would also be much simpler to use for those who don't understand regular 
expressions.

I think just supporting single-line regex replacements would go a long way.  
I'm looking at the code (I think this is the right place):

bool CDiffData::CompareWithIgnoreWS(CString s1, CString s2, DWORD dwIgnoreWS)

and imagining passing strings s1 and s2 through arbitrary user-definable regex 
filters. For example, the function could get a collection of string pairs that 
could be passed into boost::regex_replace.  For example, see 
http://stackoverflow.com/questions/13904622/replace-string-through-regex-using-b
oost-c and imagine passing in the string pair:

{"\\|[^-\\d.]*(-?[\\d.]*)\\}\\}", "$1"}

Tortoise could offer several pre-seeded string pairs, to handle single line 
comments, replacing GUIDs with the string "GUID", a common date-time format 
with the string "DATE TIME", etc.  The initial implementation would err on the 
side of a very short list of pre-seeded values.  TortoiseMerge users would 
suggest future additions.

I'll get to that in a second step. But not using boost - the c-runtime now has 
its own regex engine.

Nice, thanks.  All kinds of "show only the important differences" use cases out 
there.

This issue was updated by revision r24915.

Feature done, but UI is still missing.

This issue was updated by revision r24919.

This issue was closed by revision r24932.

Just downloaded version 1.8.8. Still no sign of an according UI option.
According to https://code.google.com/p/tortoisesvn/source/detail?r=24919 it 
should be there?!

Look at the milestone: 1.9.0
New features are never added to the stable branch.