David Wendt

> Part way through reviewing I realized that there are a number of files in the strings_udf/cpp directory that are just copied from libcudf. Where do we run into problems...

> At this point I think our options are: > > 1. pause for now and revisit later during future refactoring > 2. take a closer look at a performance...

I guess I thought that the task here was to have only a single `stod` function used by both the `parse_numeric` and the `convert-to-float` functions. It now appears we still...

> Sorry I have an off-topic comment: Should `stod` is an important enough feature that should be implemented more carefully (i.e., fully coveraged for boundary conditions) at a lower level...

This should be marked as a _breaking change_ since the public enums have been moved/renamed.

@gpucibot merge

I'm also in favor of putting these in their respective repos. Perhaps a script can be used to wrap the > I believe it would be hard to import dependencies...

> I totally see the issue - we could configure the rmm and thrust path via CMake and create the pretty-printer/load script inside `build`. Would that be an acceptable solution?...

> Considering the horrible things I have to do to try and make the deduplication work (concatenating files in CMake! `exec(open(...).read())` in Python), I'm very open to hearing alternative approaches....

> Does nobody use the regex engine from the C++ side? I am just surprised we don't have any API docs that explain the regex keywords we support. Spark uses...