Fix for #2841

[swamishiju]

Removed check for struct type and used same procedure for all empty arrays

[swamishiju]

#2841

[swamishiju]

We still need to fix print for arrays they dont seem to be working properly

#2841 (comment)

[ubaidsk]

Does this PR also fix the print issue mentioned in #2841 (comment)?

[swamishiju]

Doesn't fix the print issue i need to look into that i'm not sure rn where runtime printing is handled but it shouldn't take too long

[ubaidsk]

Ok. I think this is still good. It seems to fix some our integration tests without breaking other tests.

[swamishiju]

Does this PR also fix the print issue mentioned in #2841 (comment)?

@Shaikh-Ubaid The print issues can't be fixed until we have separated libasr because it is intended fortran behavior but not for python

Same problem with string tests also, why don't we have libasr as part of this repo now?

[ubaidsk]

Same problem with string tests also, why don't we have libasr as part of this repo now?

Going forward, the contribution workflow that we intend to use is that any changes to libasr needs to be sent to LFortran first, approved and merged there and then we just update the libasr commit in LPython. This ensures that any changes we do to LibASR in LPython are always compatible with that of LFortran. This also helps when we do LibASR sync, as now we just need to update the LibASR commit and get the lpython integration tests to pass.

[ubaidsk]

The print issues can't be fixed until we have separated libasr because it is intended fortran behavior but not for python

Could you elaborate on the intended behaviour for fortran and python, maybe sharing some small examples?

[swamishiju]

The print issues can't be fixed until we have separated libasr because it is intended fortran behavior but not for python

Could you elaborate on the intended behaviour for fortran and python, maybe sharing some small examples?

I discussed this today with Ondrej over Zulip

He said it was supposed to be handle by string_format_kind I think its probably not working

The examples I have in mind were

In fortran

Program Hello
integer, dimension(3,2) :: x
Print *, x
End Program Hello

prints

   0       0           0       0           0           0

Like in a straight line without any like formatting

This is what i think equivalent code for python would be

x:i32[3,2] = empty([3,2], dtype=int32)
print(x)

which pre-sync prints formatted

0 0 0
0 0 0 

I will look into this