command line wildcard for reading several files (without multiple -f) ?

[edpichler]

I would like to use the argument -f to add files of an entire directory (representing a year), is it possible?

[simonmichael]

I often wish for that too. It's not currently possible; at times I have used shell scripts to generate the multiple -f's.

Enhancements considered:

• -F/--files '2023/* 2024/*' - the quotes would be required, a bit clunky
• hledger ... -- 2023/* 2024/* - multiple file arguments allowed after --, also a bit clunky

What other things could we do ?

[simonmichael]

• -f '2023/*' - if -f's usual argument contains a * or a ?, treat it as a glob generating multiple -f's. (Reuses -f, but a glob character must be used, unlike -F above.)

[edpichler]

I thought something like hledger -f 2023/* -f 2024/*.dat -f index.dat would be nice.

Edit
-f 2023/** and -f 2024/**.dat where double ** could read sub dirs, maybe. A git/glob style?

[simonmichael]

That doesn't work, because the shell will expand that to

hledger -f 2023/file1 2023/file2 -f 2024/file1.dat 2024/file2.dat -f index.dat

which can't be parsed sensibly.

[edpichler]

I forgot we have this:

It will help in my current multi-file problem. I can create index files with the references I need.

[larsks]

What if you could pass a directory as the argument to -f instead of a file, and then have hledger operate on all the files in that directory? That means we don't need to deal with the problem that @simonmichael pointed out, but you still get a nice syntax for operating on batches of files:

hl import -f data/mycreditcard/

[simonmichael]

That sounds fine too, but only when you have no other kinds of file in the directory.

[simonmichael]

I think a directory often contains files that include each other. In that case, reading all files would not be desirable, as the includes will cause some of them to be read a second time, duplicating the data.

Instead, how about reading the file named like the directory, if any. Eg:

2025/
 2025.journal           <- read this file, because its base name matches the directory
 2025-accounts.journal
 2025-tags.journal
 2025-prices.journal

If no similarly-named file is found, it could default to reading the alphabetically first ? So you could designate the main file by prefixing a 0, _, or capitalising its name.

[edpichler]

@simonmichael I like the principle of "less is more".

I started this thread but I realised we can reference files from inside of it. So having an index.dat and referencing other files might be a good solution (at least for my use case) without increasing the code base and maintenance costs. Moreover, the current solution is less error prone. Undesired files could be referenced and wrong decisions could be made accepting wildcards, for instance.

[simonmichael]

+1 to that principle.

I think I like the idea of using a directory-named file as the index. More useful than a generic "index.journal" name.

(Currently doing cleanup of old journals, and I am quite often typing -f YYYY/YYYY.journal)