|
This is a good feature you are
thinking of David! Good luck and have a great weekend!
More comments
inline
below:
I'll investigate the coding
conventions. I'm pretty sure that VS.NET will "correct" the curlies by
default, but I'll try make sure that my NAnt patches adopt the alternate,
untabbed lifestyle. ;)
I have no idea what the internals of the
Framework do with that StreamReader overload. For the kilobyte-sized SQL
files I'm comparing, it seems to be minimal. I'm more than happy to
adopt an alternate route if there's a better performing
widget.
UNICODE is a goot kvestion, and somewhat
outside my general experience; I'll ask my ESL developers to help me out there
on Monday. Whatever the native StreamReader encoding-to-String is/does
is what will happen using that overload. I would lean toward throwing an
exception if the files being compared had a different encoding scheme...
Any thoughts?
Comparing file sizes may have the same
problem as you noted with UNICODE -- the sizes may match, but the files are
not identical, so that wouldn't solve my problem. One character can make
or break SQL syntax. :)
Very true about SQL.
I should
have specified that only if the filesizes were the same, should the MD5 hashes
be computed to check for a match.
If you're
trying to compare text files, then the issue of encoding comes up and you're
approach of using the StreamReader class seems to be correct because it will
give you UNICODE inside the framework even though the physical file maybe saved
as ANSI, UNICODE, UTF-8, KOR-8, etc.
There's also
the slippery slope when comparing content. Are two XML files equivalent if their
DOMs are congruent? Are caps significant in a SQL file? You'll eventually get to
the point of what if one person codes with the curly braces with the if,
and the other person doesn't. :-)
Personally,
I would lean towards binary compares, rather than text content
compares.
I hadn't considered someone attempting to
diff binaries or 20MB files. I would expect
binaries-converted-to-strings to compare same-or-not just fine... but short of
adding a hard limit on file size that may be overridden by
yet-another-property, I'd chalk it up to caveat developer on the oversized
load. Fun to see when you reach pagefile exhaustion and the o/s throws
an exception to NAnt...
isAMatch and the thatSR.Close() are
likely to be an example of why one should not be writing code at 2am local
time. Be happy to fix myself there!
I've never used the FileInfo class
before, so it's overhead is an unknown to me; I'll see if there's a less
painful way to provide some meaningful verbose logging.
I'll investigate the MD5 hash, but unless
there's some voodoo involved internal to the Framework, isn't it going to have
to read in the whole file anyway just to compute the hash for
comparison?
Yes, it'll have to read the entire file, but it won't
keep the entire file in memory like the strings do.
See:
http://msdn.microsoft.com/library/default.asp?url="">
and:
http://msdn.microsoft.com/library/default.asp?url="">
I've no idea what "alternate" streams
might be, so I'll leave that case in the corner for another
day.
Check out:
http://msdn.microsoft.com/library/default.asp?url="">
Most people
don't even now about theirs existence nor take advantage of
them.
I appreciate the feedback!!
Thanks.
You're very welcome.
-----Original
Message-----
From: Anthony
LoveFrancisco [mailto:[EMAIL PROTECTED]
Sent: Friday, September 05, 2003 6:25
PM
To: David Reed;
[EMAIL PROTECTED]
Subject: RE: [nant-dev] Who's the happy
moderator who ate my <if /> task post? :)
A couple
of comments and suggestions:
- Check
the coding guidelines for the coding style:
- opening curly braces go on the same line with the if (), else, or while
()
- By
using StreamReader, aren't you taking on on the cost of decoding the
bytestream into a character stream ?
- How do
you deal with binary files?
- What
happens if comparefile is ANSI while
difffile is
UNICODE ? I'm guessing that your comparison of thisFileContents !=
thatFileContents StreamReader may say that they are the same, even though
the files are physically different.
- What
happens when you are comparing two 20 MB files? Will there be enough memory to
hold these two strings?
- the
variable named "isAMatch" seems to be a misnomer. As I read the code, it seems
to say that the files are a match when the contents are not
equal.
- the if
statement clauses seem backwards based on "isAMatch". It says that the files
are different if isAMatch is true.
-
thatSR.Close() is outside the foreach loop. I think it should be
inside.
-
constructing a FileInfo to be used just in verbose logging seems like a
waste.
[There's
also the corner case of alternate streams on Win32. Do you need to compare
those streams as well. I think most people just ignore
this.]
- May I
suggest using the Stream class and just computing the MD5 hash of each of the
files you are comparing. A quick shortcut maybe inspecting the filesizes first
before investing into actual computing the hash.
-
Ants
-----Original
Message-----
From:
[EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of David Reed
Sent: Friday, 05 September, 2003
15:16
To:
[EMAIL PROTECTED]
Subject: [nant-dev] Who's the happy
moderator who ate my <if /> task post? :)
I'll just
paste the message below without the happy patch (since it violated some
arbitrary post length limit) so that I can get your feedback on its utility
for others. If I'm not shouted down, somebody can suggest a good
method for getting the patch out there, if the attachment
vanishes.
-----Original
Message-----
From: David
Reed
Sent: Thursday,
September 04, 2003 1:15 AM
To: NAnt Developers
([EMAIL PROTECTED])
Subject: IfTask for file contents
comparison?
Maybe this
isn't the right way to do this, but I needed to be able to compare file
contents (not just timestamps), and I couldn't find another way to do this
comparison with existing code. School me if I'm just lost.
:)
I'm proposing
the addition of an attribute called difffile to if/ifnot which will work
with the attribute comparefile (or the element comparefiles, which is just
collateral damage to me) to determine whether the contents of difffile
matches the contents of comparefile.
I'm happy with
the way it's working for me, but I'm open to suggestions that will make it
more ubiquitous if somebody's got a better idea than this patcheroo.
My documentation sux, sorry; I'm learnin' better XML commentary and NDoc
sometime next month.
|
