“CE” was just an example of a less common acronym. I think such an acronym
can be used for various purposes, so it is probably quite local in the code what it
stands for.
We also have other (I think more common) acronyms like TLI for
TargetLoweringInfo, or MF for MachineFunction.
As Krzystzof were saying “Long names should be used where
meaning needs to be conveyed, otherwise they just obfuscate the code
needlessly.”
Having to write out targetLoweringInfo instead of the short form TLI
everywhere would probably obfuscate things (lots of long lines, line splits)
rather than making things easier to read/understand.
FWIW, there is a proposal in the RFC, and the coding standard still say that
acronyms may be used. I still haven’t seen any answers about what the
“correct way” would be when using an acronym if the RFC is accepted.
I’ve only seen answers about _not_ using acronyms, but that has not part
of the RFC afaict.
My point was that the RFC need to be clear about what to do with the
acronyms if we go down that path.
I vote for treating acronyms like any other individual word written in
camel case or snake case. That way, no one has to waste cycles trying
to dodge initial or consecutive acronyms when they make sense.
For example, given acronyms FOO and BAR, the phrase "FOO BAR" becomes
"fooBar", "FooBar", or "foo_bar".
Another occasional bonus is that a dictionary of acronyms isn't
required to automatically convert arbitrary names among the forms.
Some people think UpperCamelCase is ugly for acronyms, such as "Gdb"
or "Llvm". That's the only objection I'm aware of for this approach,
and it annoys me far less than the above issues. Maybe I'm in the
minority.
Joel