I am currently making some BitcodeWriter changes that involve some refactoring, and am thinking for the Nth time that it would be much nicer to have a class instead of passing around a long list of parameters. I am thinking of biting the bullet and doing that - any objections?
I assume the reason why there is no existing class wrapping the bitcode writing process is just legacy code and nothing else?
It seems like at the least a simple class could hold the common elements used by the various bitcode writing routines, and I might want to add a specialized class to manage the ThinLTO combined index file generation.
Thanks,
Teresa
In general I’m worried about having single gigantic class that keep many data members, this goes against https://en.wikipedia.org/wiki/Single_responsibility_principle and makes it hard to track what is initialized, where, and under which condition (basically one of the reason why global variables are not welcome). The code is almost always easier to understand with small separated components (yes many places are drifting a lot in LLVM…).
Not to say that it can’t be done, just that it requires a lot of care.
I am currently making some BitcodeWriter changes that involve some refactoring, and am thinking for the Nth time that it would be much nicer to have a class instead of passing around a long list of parameters. I am thinking of biting the bullet and doing that - any objections?
I assume the reason why there is no existing class wrapping the bitcode writing process is just legacy code and nothing else?
It seems like at the least a simple class could hold the common elements used by the various bitcode writing routines, and I might want to add a specialized class to manage the ThinLTO combined index file generation.
Please do this. We'll all profit. Thanks!
Since you'll be touching all the lines anyway, it's also a good
opportunity to rename the functions to match current conventions
(i.e., start with a lower-case letter).
I don't think we should throw *all* the state in. But some of the state
is passed everywhere.
At least (off the top of my head):
- ValueEnumerator&
- BitstreamWriter&
Agree. Also the ModuleSummaryIndex, and the Module in the case of writing
bitcode for a module as opposed to a combined index (which should probably
be different derived classes). I'll try to be sensible, but you can let me
know if it is too much when I have a patch. =)
Good idea on lower-casing function names at the same time, at least for
anything that moves into the class.
Thanks,
Teresa
The problem is that most of the time, patches in isolation are perfectly fine, it is only when you step back and see the result that you figure out you reach the state of https://raw.githubusercontent.com/llvm-mirror/llvm/master/lib/Target/X86/X86ISelLowering.cpp Multiple people adding stuff after you does not help to keep the overall vision as well.
(To be clear: I don’t mean I’m opposed to what you want to do)