[PATCH] D66398: 2/2: Fix `TestDataFormatterStdList` regression

Thu Aug 22 07:42:39 PDT 2019

jankratochvil added a comment.

In D66398#1640871 <https://reviews.llvm.org/D66398#1640871>, @labath wrote:

> because these kinds of regular are horrid.

Yes, they are. So let's link with PCRE <https://www.pcre.org/> (BSD licensed).

> What was the problem with the dispatching solution?

I do not find it so easy as libc++ formatter is using static function `LibcxxStdListSyntheticFrontEndCreator` while libstdc++ is using new object instance `ScriptedSyntheticChildren`. The dispatching should be in `LibcxxStdListSyntheticFrontEndCreator` so should it store `new ScriptedSyntheticChildren` as a static variable?

And then I do not like that solution anyway. Currently the dispatching is based on type name strings so why not just do that? KISS <https://en.wiktionary.org/wiki/KISS>

> we can dispatch based on some other property. The presence of some member variable for instance?

So let's drop the type name completely and make the matching duck typed <https://en.wikipedia.org/wiki/Duck_typing>. Why to divert from how it is currently done?

Anyway sure feel free to rewrite it but the regression is fixed now.

Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66398/new/

https://reviews.llvm.org/D66398