Remove goto

[certik]

The code has about 30 occurrences of goto. We should replace it with modern Fortran's exit, cycle, return and other flow constructs.

This is very important to make people more excited to use the new code base (such as for SciPy #14), see e.g. this comment:

I’m not sure if that’s what actually modern Fortran looks like - I had expected not, because there’s still a lot of goto's in the code (which is bad).

[milancurcic]

FWIW, the use of goto in the current minpack is fine. Most of them are to avoid repeating the error handling boilerplate. I would otherwise recommend not changing it, but I understand that from a cultural perspective it's important to change it. If it's important for SciPy for goto to go, let's do it.

[jacobwilliams]

GOTOs gotta go. Perception is very important. We can refactor and not duplicate code.

[certik]

Perception is everything for minpack, it's our "poster child". Let's figure out how to do the error handling correctly without goto, and without losing any performance and without duplicating code.

[rgommers]

FWIW, the comment @certik linked was mine. Please feel free to take anything I say with a few grains of salt, since I really don't know much about Fortran. C code has some goto's sometimes as well, but they're usually goto error or goto exit; to me that is very different from goto 300, since I don't have to read the lines of code to figure out what the purpose of the goto is.

[certik]

@rgommers actually, what you say is very important, as you are our ideal "customer". You are not a Fortran person, but you are in charge of quite a lot of Fortran code in SciPy and you are not against Fortran if we can fix all the very real technical problems with Fortran (in the broader sense). So we really appreciate your frank feedback.

[arjenmarkus]

I have not looked at the code yet, but can we not use a named BLOCK construct, so that most of the GOTOs (if not all), are replaced by "exit ordinary_handling" or the like? At the end of the block you then get all the error handling. Op ma 6 jun. 2022 om 16:24 schreef Ondřej Čertík ***@***.***>:

…

@rgommers <https://github.com/rgommers> actually, what you say is very important, as you are our ideal "customer". You are not a Fortran person, but you are in charge of quite a lot of Fortran code in SciPy and you are not against Fortran *if* we can fix all the very real technical problems with Fortran (in the broader sense). So we really appreciate your frank feedback. — Reply to this email directly, view it on GitHub <#74 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAN6YR5QOLKENF4WAXAON5TVNYCZJANCNFSM5X7KAQLQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jacobwilliams]

See the PR #75. So far, just for hybrd, but it's using this approach. added a block and two do loops.