[clang-tidy] New check: Use if-initializer to reduce scope of a variable

[SunBlack]

Over the years, we have accumulated a lot of codes like this one:

const auto it = myMap.find(value);
if (it != myMap.end())
{
  // do sth with the iterator
}

Since C++17 there are if-initializers that make it easier to see whether the variable is only needed in one block or whether it will be needed again later.

So you can write sth. like this:

if (const auto it = myMap.find(value);
    it != myMap.end())
{
  // do sth with the iterator
}

Maybe this could also be part of a bigger check to reduce the variable scope in general (similar to to the variableScope check in Cppcheck).

[llvmbot]

@llvm/issue-subscribers-clang-tidy

Author: SunBlack (SunBlack)

Over the years, we have accumulated a lot of codes like this one: ```cpp const auto it = myMap.find(value); if (it != myMap.end()) { // do sth with the iterator } ```

Since C++17 there are if-initializers that make it easier to see whether the variable is only needed in one block or whether it will be needed again later.

So you can write sth. like this:

if (const auto it = myMap.find(value);
    it != myMap.end())
{
  // do sth with the iterator
}

Maybe this could also be part of a bigger check to reduce the variable scope in general (similar to to the variableScope check in Cppcheck).

[EugeneZelenko]

switch statement should be considered too.

[carlosgalvezp]

Sounds good, I think this fits in modernize.

A separate check could be done for reducing scope of other variables in general.

[EugeneZelenko]

Sounds good, I think this fits in modernize.

A separate check could be done for reducing scope of other variables in general.

This is #22366.

[vbvictor]

Maybe wild suggestion, but could we close old Bugzilla issues in favor of new ones? Assuming all unique additional context from old issue will be transferred to new one.

IMO, a new contributor won't likely try to implement something from 2014 (if he ever finds that issue). Last 10 pages of clang-tidy issues (primarily from Bugzilla) feels like a graveyard.

[EugeneZelenko]

Maybe wild suggestion, but could we close old Bugzilla issues in favor of new ones? Assuming all unique additional context from old issue will be transferred to new one.

IMO, a new contributor won't likely try to implement something from 2014 (if he ever finds that issue). Last 10 pages of clang-tidy issues (primarily from Bugzilla) feels like a graveyard.

I don't think that keeping new issue will help situation too much because if may be buried in very similar fashion. Obviously keywords search is useful.

Current situation just indicates that number of ideas are much bigger then developers' time to implement them and it persist since start of Clang-Tidy :-(