DM> create a program with the following spec:
DM>
DM> 1. Written in pure, portable C or C++
DM> 2. Can produce a "diff", or "difference",

I've got something like that.... It's only 160 lines...

DM> 3. No line, whether context, addition or deletion, should exceed 75
DM> 4. Can have a maximum number of lines per output file.  See CSplit.

Why not just use csplit to "wrap" the output. (pun intended)

DM> Comparator: This will need some sort of memory to remember a number of
DM> lines  to produce the context for output.

you know, I just made _extensive_ use of fseek() :) that's why I run a disk
cache :) setting the buffers to 32K helps too. (see my code)

DM> Anyone want to take a crack at it?  :-)

I'll post some source, you should be able to add the features you want fairly
easily. mine does line numbers, but by adding a ring-buffer to hold the file
offsets of the previous lines you'd be able to fseek back to a few lines
before the first non-matchine line... while I was de-bugging it I used all
sorts of different ouput code, all the output code is "centralised" in one
location, so it's easy to change it to produce different output styles....

And now the bad news...

How much context is sufficient? there may be more than one place where a line
or sequence of lines occurrs in a program, managing to get a sufficient amount
of context without being wastefull seems tobe a matter of re-scanning the file
from the location of the last "edit" to check how much context is needed.

The other problem with my code is that it never gives up, if you give it 2
1000 line files with no matches, it'll make sure, by comparing every line in
each file... with every line in the other file...  that could take a while.

I'll post what I have now, and let everyone here "featurise" it for me :)
(Code in next messages)

Bye.

--- EzyQwk V1.20 01fa018d