cppunit-devel

[Cppunit-devel] regular expression match

[Volker B. <vbo...@te...>]

Hi,

for quite a while now it has been a plan of mine to add some sort of
regular expression support to cppunit. Whenever it gets to comparing
strings it would often help to use regular expressions instead of
exact string match.

Often dynamic information is the problem. An example: I want to check
the output of a logging server. Every message should have this format:

  <msg. sequence no.> <process id> <date> <time> <message>

e.g.

  2 15337 2002-04-05 09:44 server started

With current cppunit I use shell scripts to check the output. With regular
expressions it would be much easier, e.g.

  CPPUNIT_ASSERT_REGEX_MATCH("^\d+ \d+ 20\d{2}-\d{2}-\d{2}"
    "\d{2}:\d{2} server started", logline);

(with perl style \d :=3D [0-9]).

But there might be some issues with using regular expressions in C++:

 - Although regular expression handling is part of the POSIX standard
   I don't know how portable it is.
 - The regex interface is somewhat more complicated than a simple
   string equality. There are a lot of switches and options
   controlling the type of match. But it should be *really simple* to
   use in cppunit.
 - There are some concurring regex implementation. On Debian Linux
   e.g. there are at least 5: glibc (libc/POSIX), libregex (GNU),
   libregexx (C++ wrapper for libregex), libboost (C++, maintained by
   well known Steve M. Robbins...), libpcre (perl style regex)

I think, we should have at least one simple regex-assert, at least on
some platforms that supports regexps.

Perhaps we could have a requirement like this 'if you want to use
regexps with cppunit you have to install libXXX otherwise they will be
disabled' (the way cppunit deals with doxygen). I think this has some
advantages:

  - One could use the 'best suited' library, e.g. a C++-library.
  - There might be less portability issues.

Comments?
Are there even any existing implementations of regex-asserts already?

Best regards,
Volker

(BTW. regexps are still less than one could think of: In the example
above there is no way to ensure that the timestamp is a valid
date...)
--=20
Volker Boerchers <vbo...@te...>
System Engineer
TECON Systems AG, Perlengraben 2, 50676 K=F6ln, http://www.tecon.de
Tel: +49-221-92007-55, Fax: +49-221-92007-77

Re: [Cppunit-devel] regular expression match

[Steve M. R. <ste...@vi...>]

Hi,

I haven't any real opinion on adding regular expression matching
to CppUnit.  I presume one could always call their own favourite
regex matcher wrapped up in CPPUNIT_ASSERT without too much trouble.

However, since you mentioned the BOOST libraries, I wonder if you have
looked at their implementations of concept checking and testing?  The
web page (www.boost.org) indicates that there is not yet any unit test
library.  The BOOST code is pretty well designed and the mailing list
is full of very bright people.  So it would be a boon to have CppUnit
involved with them.

I haven't looked at any of this stuff, myself.  Has anyone else?

-Steve

-- 
by Rocket to the Moon,
by Airplane to the Rocket,
by Taxi to the Airport,
by Frontdoor to the Taxi,
by throwing back the blanket and laying down the legs ...
- They Might Be Giants

Re: [Cppunit-devel] regular expression match

[Duane M. <dua...@ma...>]

--- At Wed, 10 Apr 2002 08:45:39 -0400, Steve M. Robbins wrote:

>Hi,
>
>I haven't any real opinion on adding regular expression matching
>to CppUnit.  I presume one could always call their own favourite
>regex matcher wrapped up in CPPUNIT_ASSERT without too much trouble.
>
>However, since you mentioned the BOOST libraries, I wonder if you have
>looked at their implementations of concept checking and testing?  The
>web page (www.boost.org) indicates that there is not yet any unit test
>library.  The BOOST code is pretty well designed and the mailing list
>is full of very bright people.  So it would be a boon to have CppUnit
>involved with them.
>
>I haven't looked at any of this stuff, myself.  Has anyone else?

I am a subscriber of the boost users list <mailto:boost-users-
sub...@ya...> or <http://www.boost.org>. 

Boost has a C++ regular expression interface as well as a unit test
framework. 

The unit test framework has no relationship to CppUnit. I have mentioned
it to the developer of the Boost framework with no real response. The
present unit testing is buried in boost, I believe to test some of boost.
But the intention is that it will eventually be something like CppUnit.

I find this very disappointing as CppUnit is much farther along. I
imagine that CppUnit could also benefit greatly from the input of the
Boost group.

As I understand it, the boost regular expression library is meant to be
Perl compatible (although why not posix I am not sure). I believe boost
may be working towards getting the regular expression library into the
C++ standard library.

Be aware that I have used neither of these libraries.

 ...Duane

-- 
"If tyranny and oppression come to this land, it will be in the
guise of fighting a foreign enemy."              - James Madison

RE: [Cppunit-devel] regular expression match

[Arlo B. <arl...@ho...>]

I've been using the boost libraries for a while, including their regexp
stuff. If you're looking for a regexp library, I'd recommend Boost's.

In particular, it works well with STLPort, which means that the list of
currently supported platforms includes all Windows, Mac OS X, Mac OS
6.something +, a few dozen Linux distros, most Unixes, Amiga, PS2, XBox, and
so on. You don't have to worry about platform availability, although you do
have to ensure that the developer installs the STLPort and Boost libraries.

Likewise, he boost regex was built with Unicode in mind. It performs all
matching using char_traits, so should internationalize as well as the rest
of the STL (YMMV). It supports most digraphs, and matches umlauted,
accented, and otherwise modified characters correctly. I expect it to work
with the Asian, Arabic and Hebrew languages, but don't really know.

Additionally, it uses a pretty full regex syntax, similar to Perl's. I
haven't found much missing, although the regex strings end up as even more
line noise than they do in Perl, because you have to escape the backslash
character past the C++ processor any time that you put in an escape
sequence.

There is a learning curve on the boost regexp - it tries to expose a
completely generalized interface, which leads to complexity. As most STL
things, it defines algorithms that operates on ranges, which means that you
can trivially use it to match not only on a string, but also on any
container, on files directly, or an a wrapped stream (it requires on
bi-directional iterators, so you can't just match on cin directly). However,
this adds some complexity to its use - there are lots of results classes,
templates, and syntactical garbage to keep track of.

For the purposes of CppUnit, you are really only looking for a match / no
match result, so you could easily wrap it all in a functor which returns a
bool. Then a simple macro to assert on that, and voila. Note, however, the
use of you in that sentence. Sorry guys, but I've got a ship date coming up
at work, and won't be able to get around to just writing this for a while.

Another option would be to have a family of such functors that use the
various regex libraries, and have CppUnit select one based on what libs are
installed on the developer's machine (either statically or dynamically). I
don't know how feasible it would be to check for support on other platforms
(haven't tried it), but it's trivial to tell whether Boost & STLPort are
installed on a Window box, as they are both DLLs. In any case, static config
is simple enough.

I agree that it would be good to have these two projects become involved
with each other. Note, however, that CppUnit isn't really architected in the
way that they're planning to build the test library. Probably both sides
will change on this one - as Steve pointed out, they don't have an
implementation.

Arlo

> -----Original Message-----
> From: cpp...@li...
>
> I haven't any real opinion on adding regular expression matching
> to CppUnit.  I presume one could always call their own favourite
> regex matcher wrapped up in CPPUNIT_ASSERT without too much trouble.
>
> However, since you mentioned the BOOST libraries, I wonder if you have
> looked at their implementations of concept checking and testing?  The
> web page (www.boost.org) indicates that there is not yet any unit test
> library.  The BOOST code is pretty well designed and the mailing list
> is full of very bright people.  So it would be a boon to have CppUnit
> involved with them.
>
> I haven't looked at any of this stuff, myself.  Has anyone else?

RE: [Cppunit-devel] regular expression match

[Volker B. <vbo...@te...>]

On Wed, 10 Apr 2002, Arlo Belshee wrote:

> I've been using the boost libraries for a while, including their regexp
> stuff. If you're looking for a regexp library, I'd recommend Boost's.

I stumbled across the boost project searching for a regexp library but
was pretty amazed that I never heard about this large and nice thing
before.

> In particular, it works well with STLPort, which means that the list of
> currently supported platforms includes all Windows, Mac OS X, Mac OS
> 6.something +, a few dozen Linux distros, most Unixes, Amiga, PS2, XBox, =
and
> so on. You don't have to worry about platform availability, although you =
do
> have to ensure that the developer installs the STLPort and Boost librarie=
s.
>=20
> Likewise, he boost regex was built with Unicode in mind. It performs all
> matching using char_traits, so should internationalize as well as the res=
t
> of the STL (YMMV). It supports most digraphs, and matches umlauted,
> accented, and otherwise modified characters correctly. I expect it to wor=
k
> with the Asian, Arabic and Hebrew languages, but don't really know.
>=20
> Additionally, it uses a pretty full regex syntax, similar to Perl's. I
> haven't found much missing, although the regex strings end up as even mor=
e
> line noise than they do in Perl, because you have to escape the backslash
> character past the C++ processor any time that you put in an escape
> sequence.

Thank you for this evaluation!

When I find the time I will implement an asserter using the
boost-library and try to add the necessary autoconf stuff. As long as
one makes the library selectable via configure this dependency should
be acceptable I think. Regex matching would be an add-on and no basic
functionality would depend on it.

Regards,
Volker
--=20
Volker Boerchers <vbo...@te...>
System Engineer
TECON Systems AG, Perlengraben 2, 50676 K=F6ln, http://www.tecon.de
Tel: +49-221-92007-55, Fax: +49-221-92007-77

Re: [Cppunit-devel] regular expression match

[Baptiste L. <gai...@fr...>]

----- Original Message -----
From: "Steve M. Robbins" <ste...@vi...>
To: "CppUnit Development" <cpp...@li...>
Sent: Wednesday, April 10, 2002 2:45 PM
Subject: Re: [Cppunit-devel] regular expression match


> Hi,
>
> I haven't any real opinion on adding regular expression matching
> to CppUnit.  I presume one could always call their own favourite
> regex matcher wrapped up in CPPUNIT_ASSERT without too much trouble.
>
> However, since you mentioned the BOOST libraries, I wonder if you have
> looked at their implementations of concept checking and testing?  The
> web page (www.boost.org) indicates that there is not yet any unit test
> library.  The BOOST code is pretty well designed and the mailing list
> is full of very bright people.  So it would be a boon to have CppUnit
> involved with them.
>
> I haven't looked at any of this stuff, myself.  Has anyone else?

I have used some of boost component (smart-pointer and regular expression).
There is lot of interesting thing, but not always supported by all compiler
(tuple is pretty much useless on VC++).

I haven't looked to deep into the unit test (I tried to give it a look, but
they are usually a large bunch of lines, pretty much unreadable for me.).
Still, they test for memory leak.

Baptiste.

Re: [Cppunit-devel] regular expression match

[Baptiste L. <gai...@fr...>]

You can easily add your own assertion using Asserter. Look up the
documentation menu 'module': creating you assertion.

At the current time, I don't wish to introduce dependency uppon boost
library (weird build system, and not has more portability problems than
cppunit. Still a great lib, though.).

Baptiste.

----- Original Message -----
From: "Volker Boerchers" <vbo...@te...>
To: <cpp...@li...>
Sent: Wednesday, April 10, 2002 9:04 AM
Subject: [Cppunit-devel] regular expression match


Hi,

for quite a while now it has been a plan of mine to add some sort of
regular expression support to cppunit. Whenever it gets to comparing
strings it would often help to use regular expressions instead of
exact string match.

Often dynamic information is the problem. An example: I want to check
the output of a logging server. Every message should have this format:

  <msg. sequence no.> <process id> <date> <time> <message>

e.g.

  2 15337 2002-04-05 09:44 server started

With current cppunit I use shell scripts to check the output. With regular
expressions it would be much easier, e.g.

  CPPUNIT_ASSERT_REGEX_MATCH("^\d+ \d+ 20\d{2}-\d{2}-\d{2}"
    "\d{2}:\d{2} server started", logline);

(with perl style \d := [0-9]).

But there might be some issues with using regular expressions in C++:

 - Although regular expression handling is part of the POSIX standard
   I don't know how portable it is.
 - The regex interface is somewhat more complicated than a simple
   string equality. There are a lot of switches and options
   controlling the type of match. But it should be *really simple* to
   use in cppunit.
 - There are some concurring regex implementation. On Debian Linux
   e.g. there are at least 5: glibc (libc/POSIX), libregex (GNU),
   libregexx (C++ wrapper for libregex), libboost (C++, maintained by
   well known Steve M. Robbins...), libpcre (perl style regex)

I think, we should have at least one simple regex-assert, at least on
some platforms that supports regexps.

Perhaps we could have a requirement like this 'if you want to use
regexps with cppunit you have to install libXXX otherwise they will be
disabled' (the way cppunit deals with doxygen). I think this has some
advantages:

  - One could use the 'best suited' library, e.g. a C++-library.
  - There might be less portability issues.

Comments?
Are there even any existing implementations of regex-asserts already?

Best regards,
Volker

(BTW. regexps are still less than one could think of: In the example
above there is no way to ensure that the timestamp is a valid
date...)
--
Volker Boerchers <vbo...@te...>
System Engineer
TECON Systems AG, Perlengraben 2, 50676 Köln, http://www.tecon.de
Tel: +49-221-92007-55, Fax: +49-221-92007-77


_______________________________________________
Cppunit-devel mailing list
Cpp...@li...
https://lists.sourceforge.net/lists/listinfo/cppunit-devel