remove cppname

[jmitrevs]

Variable name and cppname are often used interchangeably in the code, not consistently. My understanding is that cppname was meant to be a cleaned up C-legal version of name. Instead of looking at each case where name is used and trying to see if it should be cppname and vice versa, it probably makes sense to just get rid of cppname altogether, and just use name everywhere.

One question is whether we want to include this name cleaning in the Variable.__init__ function:

self.name = re.sub(r'\W|^(?=\d)','_', self.name)

I have not included it. The standard pytests all pass without it.

[vloncar]

Cleaning the name at that point would in fact break the model if a change actually happened. cppname dates back to when the output variable names followed the layer names, so naming a layer "my layer" would generate invalid C++. Since we switched to using layer{index}_out as the name of the output variable in generated C++, this is no longer relevant so cppname can be removed. I forgot to do it as part of the backend restructuring. Thanks for this.

[jmitrevs]

For Quartus we rename name for structure variables (the input and output). Is that safe?

[jmitrevs]

I would like if possible to have this merged fairly soon since it effects new developments, which have to consider this PR.

[vloncar]

This looks fine to me, I tried it on my Vivado developments and nothing breaks. But Quartus worries me a bit. I remember (ab)using cppname there for something. @bo3z Can you try this out and if there are no issues we merge it ASAP?

[bo3z]

I've tested this a couple of days ago, all the tests pass and synthesis is successful. I'm happy to merge this. Once merged I will have to rebase and reflect these changes in the streaming PR, but I would rather merge #562 first.

…

On Tue, 12 Jul 2022, 18:41 vloncar, ***@***.***> wrote: This looks fine to me, I tried it on my Vivado developments and nothing breaks. But Quartus worries me a bit. I remember (ab)using cppname there for something. @bo3z <https://github.com/bo3z> Can you try this out and if there are no issues we merge it ASAP? — Reply to this email directly, view it on GitHub <#562 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOIYLWZJOQUBBE3HBKA3GDDVTWN5DANCNFSM5XV72MSA> . You are receiving this because you were mentioned.Message ID: ***@***.***>