Drop uppercase F32 and F64 float number suffixes

[rhysd]

With Crystal 0.32.1,

# OK
0f32

# OK
1f32

# ERROR
0F32

# OK
1F32

This PR fixes 0F32 not to cause an error. EDIT: We decided to drop uppercase suffixes. Please see following discussion.

[rhysd]

Hmm, I'm not able to locate which test case fails on CircleCI.

https://circleci.com/gh/crystal-lang/crystal/38220?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-build-link

Error while trying to determine if a stack overflow has occurred. Probable memory corrpution

Invalid memory access (signal 11) at address 0x40

[0x5621f5b60026] *CallStack::print_backtrace:Int32 +118

[0x5621f55889cc] __crystal_sigfault_handler +348

[0x5621f72a9fe4] sigfault_handler +40

[0x7f349bd970d4] ???

[straight-shoota]

As far as I know, this is intended. Type suffixes use only lower case characters. That also affects all similar identifiers. What makes you think this needs to support upper case?

test_alpine failing is not related to this PR, you can ignore it.

[asterite]

As far as I know, this is intended. Type suffixes use only lower case characters.

It seems 2F32 and 4F64 parse well, but using capital I for integer does not. So I think we should actually drop the capital F in those cases.

[rhysd]

@straight-shoota @asterite

Actually 4.0F32 is accepted. I'm not understanding the intention, but lexer implementation explicitly accepts uppercase.

crystal/src/compiler/crystal/syntax/lexer.cr

Line 1470 in 5519fbd

when 'f', 'F'

So I thought 'F' should be accepted in the case of 0F32.

I agree that uppercase suffix looks weird. There is no test for F32 and F64 suffixes. So I feel it's better to drop uppercase suffix.

But it is a breaking change. Should we do it? I'm sorry that I'm not understanding how Crystal dev team currently makes a decision around breaking changes. If we should do, I'll update this PR to drop the uppercase F32 and F64 suffixes. Modification would be easy.

[asterite]

Yes, I think we should drop F, even if it's a breaking change. Probably not many use it or know about it, and fixing it is easy.

[rhysd]

OK, I'll update this branch tonight.

[rhysd]

@straight-shoota @asterite

I modified this PR for dropping the uppercase suffixes. And I also added small tests for invalid float number suffixes. Would you review this?

[straight-shoota]

I'm almost sure upper case characters are unlikely to be in use somewhere. But we should at least consider a deprecation path, either updating the formatter first or providing a deprecation message when an upper case character is encountered.
IMO it's not necessary in this case because it's completely undocumented and probably not used.

[bcardiff]

If some shard is using F, it can be changed to f and publish a patch release if needed. That will also work in previous crystal versions.

I'm fine with merging this right away.

[rhysd]

CI fails due to #8782 (comment) but it seems not related to changes in this PR since in other PR it also fails with the same error. It's weird that #8781 does not cause the error, tho.