Wilfred Hughes

Just to be explicit here: what path is it printing, and what would be the ideal output?

OK, I have a minimal standalone repro. ```rust trait VCipherSuite {} trait CipherSuite where OprfHash: Hash, { } type Bar = ::Hash; type OprfHash = ::Hash; impl Foo { fn...

Thanks for the report. Could you share an example of files where the identifiers differ by case, not just the keywords?

Note that PHP also has case insensitive keywords, but some identifiers are case sensitive: https://stackoverflow.com/questions/20624257/are-php-keywords-case-sensitive

This code looks sensible to me, but I'm slightly nervous about changing C++ in core DB code when I have less context on the C++ components. I'm trying to find...

This doesn't look great, which is why the original approach was chosen. 

The file `Protos/Tests/SwiftProtobufTests/generated_swift_names_fields.proto` in https://github.com/apple/swift-protobuf/commit/e271b13e96c351084229eb0a9a55290b4002ba01#diff-3bd8bc669a3a912d4a8381aec5bc7704ddb1b48e9f054b09a76d758dac097cea seems to be hitting the graph limit and might be a good test case. See also #629 and #623.

https://github.com/apple/swift-protobuf/commit/60153e3ee8adf563dac27c19d1c680ba5c723312#diff-d9afc38bdc272c3d9e8f79273b50970fd122303e5b04ffb82fc5c758091c372d is another interesting test case, only a few lines changing in a large file but hitting the graph limit.

I don't think there's any systematic way for difftastic to know what the embedded language is here. For HTML we can detect JS or CSS embedded snippets, but arbitrary string...

Our HSL CI fails on this diff, I presume because you end up with `` which doesn't allow extension at all. Perhaps we should just not seal it in OSS?