Python bind for osh

[KingMob]

No description provided.

[andychu]

I ran the spec tests locally and got this -

case    osh     bash
Ignoring osh-cpp failure: 0 test bind (w/out flags) for adding bindings to readline fns
0       FAIL    ok      test bind (w/out flags) for adding bindings to readline fns
        0/5 ok
Ignoring osh-cpp failure: 1 test bind -r for removing bindings
1       FAIL    ok      test bind -r for removing bindings
        0/5 ok
Ignoring osh-cpp failure: 2 test bind -x for setting bindings to custom shell functions
2       FAIL    ok      test bind -x for setting bindings to custom shell functions
        0/5 ok
3       ok      ok      test bind -u for unsetting all bindings to a fn
4       ok      ok      test bind -q for querying bindings to a fn
5       ok      ok      test bind -m for setting bindings in specific keymaps
6       ok      ok      test bind -f for setting bindings from an inputrc init file

Is that what is expected to work?

The code looks reasonable, and this seems like good progress!

---

If so, I can help fix the C++ issues

[KingMob]

Is that what is expected to work?

Yes, this is what I expect. The failing tests don't have code implemented for them yet.

[andychu]

I pushed a couple minor commits to pass type checking and translation

So make sure to PULL before editing any code

---

It looks like we can get past the rest of the C++ build errors without much work too -- I will look at that

This looks VERY promising, thank you!

[andychu]

OK I pushed some stubs for C++ -- wasn't that bad, although there was one mycpp workaround necessary

Let's see what CI results looks like!

[andychu]

Oh weird there are some line_input.c compile errors that don't happen on my machine either

http://op.oilshell.org/uuu/github-jobs/8226/interactive.wwz/_tmp/soil/logs/py-all-and-ninja.txt

Maybe this is due to older GNU readline versions?

Maybe bash added a new flag recently that depends on a new version or something?

[KingMob]

Hmm, I haven't seen that error either. The rl_function_of_keyseq_len fn exists in my bash. I see it in lib/readline/readline.h and lib/readline/bind.c.

Looks like that particular fn was added six years ago for bash 5.0. I'm guessing python's readline wrapper uses an older version as its basis? Bash 4.4 used rl_bind_keyseq to do removal by passing a null ptr for the fn to bind to.

I'll look into what precise version python's readline was based off of. Using that as a basis is probably easier; bash/readline is old, but it's still changing, and I don't think I want to bring in the latest readline just for this. (Honestly, I'm not sure you want to use readline over a more modern lib, even without the license issue.)