On Wed, May 22, 2013 at 11:41 PM, Richard Smith <[email protected]>wrote:
> On Wed, May 22, 2013 at 9:20 AM, Jordan Rose <[email protected]>wrote: > >> >> You can tokenize without a lexer, but the line/caret combo might >> actually be better. Nice work! >> >> >> ================ >> Comment at: lib/Basic/SourceLocation.cpp:70-71 >> @@ +69,4 @@ >> + StringRef Buffer = SM.getBufferData(LocInfo.first); >> + int First = std::max<int>(Buffer.rfind('\n', LocInfo.second) + 1, 0); >> + int Last = std::min<int>(Buffer.find('\n', LocInfo.second), >> Buffer.size()); >> + OS << "|" << Buffer.substr(First, Last - First) << "|\n" << IndentStr >> ---------------- >> Not super-happy about relying on the value of StringRef::npos here, but >> okay. >> > > Aren't these min/max calls both redundant? Assuming we're not dealing with > buffers longer than 2^31 characters, they're equivalent to: > > int First = Buffer.rfind('\n', LocInfo.second) + 1; // because this is > never negative > int Last = Buffer.find('\n', LocInfo.second); // because this is never >= > Buffer.size() after conversion to int > > These are both exactly what was wanted, but should be of type size_t (== > StringRef::size_type). > They were left-overs that I didn't actually want to rely on, but forgot to clean up :) I think the latest version is easier to understand, ptal...
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
