Josh Levy-Kramer

Ow do you mean to add this to the template comment in the source file? But would that trigger a rebuild as the template is only read if it's determined...

Thanks - the above works but it would be better if the dependency file was generated on the fly. This code doesn't seem to work: ``` import hashlib from pathlib...

Ok I tested it here: `scratch.py`: ``` import os from cppimport.importer import is_build_needed, setup_module_data from cppimport.templating import run_templating cpp_code = """// cppimport #include int add(int i, int j) { return...

This would be a really nice feature. Otherwise your just repeating yourself in the docstring

I've made a pull request #109 which adds a decorator which parses the Sphinx docstring to add type and description information. It hasn't got tests yet but wanted it out...

Hi, thanks. What refactoring did you have in mind? What do you mean by a non-KISS decorator?

## Data structure > It is better to avoid allocating strings for tokens in each pre-processing step and instead use a slice of the original document. Performance depends very strongly...

> +1 generally on that. The initial reason I didn't go with it initially, I think, was that the signature of the transform method wouldn't be exactly the same between...

[I ran a benchmark](https://github.com/joshlk/vtext/commit/870e1934c0ad13b843bebb8f069b7aebe6f93833) comparing using `Box` and `impl Iterator`. The results are: ``` test dyn_pipeline ... bench: 206,121,099 ns/iter (+/- 53,332,452) test impl_pipeline ... bench: 217,948,970 ns/iter (+/- 157,269,946)...

Another thought regarding pipeline design. I am experimenting with creating Rust functions that input and output iterators that can be linked together in Python. For example in the below code:...