@dtikhonov I was considering how to help others trying to compare the QIF files generated from the output of H2O, and noticed that adding --diff option to sort-qif.pl might be a win for all of us. WDYT?

I think that in this case, worse is better (after all, I do live in New Jersey). Placing the logic in the program hides how the comparison works. People not familiar with Perl -- alas! those are in the violently loud and proud majority now -- will be lost.

In other words, I do not agree with the premise; comparing QIF files is not difficult.

A different problem is the --diff argument itself: is the program supposed to sort or to diff? The UNIX philosophy is to do one thing and do it well. Perhaps make a different program named diff-qifs.pl?

Nevertheless, so long as the sorting behavior of sort-qifs.pl remains unchanged, I am not against you adding another option to it. I would like, however, to keep the current diffing instructions in the README; the "easier" way to diff should be listed as an alternative.

Oh, and the diff mode won't work on systems without /dev/fd/fileno.

@dtikhonov Thank you for your comments.

I will change the approach to creating a script file dedicated to taking the diff.

The question is if we can assume that the diff script will exist next to sort-qifs.pl in the same directory, but I think we can assume that. Then, it becomes simply a matter of doing pathname($0) . "/sort-qifs.pl" in the diff script.

Oh, and the diff mode won't work on systems without /dev/fd/fileno.

Yeah I can change the exec arguments to bash -c exec diff /dev/fd/x /dev/fd/y, so that bash's emulation of /dev/fd will kick in on systems without /dev/fd. Then, it would work on any system with bash being installed.

IIUC, current recommendation of using <(...) requires the running shell to be bash. Changing the requirement to "either bash is installed or /dev/fd is available" will be an improvement :-)

We can make the README more portable by using temporary files.