FFINDEX_APPLY_MPI:(1)

✨ FFINDEX_APPLY_MPI:(1) ✨

okay bestie, gather round, because this one is the manual page for ffindex_apply_mpi: — the little legend that applies a program to each FFindex entry, but make it mpi enhanced 🔥. yep. distributed. parallel. main character energy.

---

💖 NAME 💖

ffindex_apply_mpi: — apply a program to each FFindex entry (mpi enhanced) 🦋

---

🌸 DESCRIPTION 🌸

so the very first thing this thing ever says to you, if you show up with nothing, is this:

ERROR Not enough arguments.

not me getting read to filth by a binary at 2am 😭 iconic honestly.

and then it hits you with the whole syllabus, literally:

USAGE: ffindex_apply_mpi [-m] [-q] [-l LOG_FILE] -d DATA_FILENAME_OUT -i
INDEX_FILENAME_OUT DATA_FILENAME INDEX_FILENAME -- PROGRAM
[PROGRAM_ARGS]*

so like. that's the shape of it. here's what the flags do, no gatekeeping:

• -l LOG_FILE — a log file that's lowkey structured like an FFindex file, except it also has the return value of the apply crammed in there. which means you can totally process it and use it as an FFindex file. for example: grep " 1$" foo.log | cut -f 1-3 > foo-special.ffindex. where maybe the ffindex_apply_mpi ran with -- perl -e 'return 1 if(/^SPECIAL/)' 🎀
• -m — do NOT merge the FFindex parts that the different MPI processes made. this one is for the big huge giant MPI jobs where the merge time might get accounted. so yeah, sometimes you're just not merging. slay 💅
• -d — the FFDATA output file. stdout of the program will be captured. caught. held. 🫶
• -i — the FFINDEX output file, and the entries get the exact same name as the input entry. no renaming drama. ✨

and, credit where credit is due:

Designed and implemented by Andreas Hauser <andy@splashground.de>.

then, and I cannot stress this enough bestie, the whole thing goes off the rails and does it again:

debian/tmp/usr/bin/ffindex_apply_mpi: Invalid argument

EXCEPTION ffindex_apply_mpi in rank: 0

ERROR Not enough arguments.

oop 💀 the full USAGE: line appears again:

USAGE: ffindex_apply_mpi [-m] [-q] [-l LOG_FILE] -d DATA_FILENAME_OUT -i
INDEX_FILENAME_OUT DATA_FILENAME INDEX_FILENAME -- PROGRAM
[PROGRAM_ARGS]*

and the flags, again, for the people in the back:

• -l LOG_FILE — like an FFindex file but with the return value of apply added. this file can be processed and used as an FFindex file. e.g. grep " 1$" foo.log | cut -f 1-3 > foo-special.ffindex. where maybe the ffindex_apply_mpi ran with -- perl -e 'return 1 if(/^SPECIAL/)'
• -m — do not merge the FFindex parts generated by the different MPI processes. this is useful for large MPI Jobs where merge time might be accounted.
• -d — FFDATA output file, stdout of program will be captured.
• -i — FFINDEX output file, entries get the same name as input entry.

and again, the icon himself:

Designed and implemented by Andreas Hauser <andy@splashground.de>.

and again, the chaos:

debian/tmp/usr/bin/ffindex_apply_mpi: Invalid argument

EXCEPTION ffindex_apply_mpi in rank: 0

it's giving recursive. it's giving "the man page ate its own tail and I'm obsessed." 🌀

---

🐛 BUGS 🐛

user feedback is welcome!! especially bugs, performance issues, and last but not least the convenience of the programs and the API. tell them everything. they want to know. 💖

email Andreas Hauser hauser@genzentrum.lmu.de. 📧

---

and that's the whole vibe, bestie. this page lands courtesy of ffindex_apply_mpi: invalid opt... from January 2026, filed under FFINDEX_APPLY_MPI:(1) — go forth and apply things, mpi enhanced, no cap. 👑🌈