Improve performance a lot

[ariasuni]

• Smarter use of buffers
• Read and write to file instead of standard input and output

Benchmarks on my computer on https://github.com/BoboTiG/ebook-reader-dict/releases/tag/fr

Original code: ⩾9s
This code: ⩾2s

Every change has been tested before/after with hyperfine and I checked that the produced files are exactly the same for several (2) dictionaries.

Ideally, I would like to be able to bypass dictfmt in the Plato script and produce directly a DICT.org file but I have trouble finding a good documentation about the format.

[baskerville]

We could even go further and not have to convert from StarDict by adding support for this format in the dictionary module. You might also be interested in this pull request.

[ariasuni]

Sadly, I don’t think I will have time in the foreseeable future to work on that. Still, was fun to work on and an immediate improvement over what’s already existing.

I opened an issue on ebook-reader-dict’s repo so that they could provide directly the correct format for Plato.

[ariasuni]

I’m not sure if you’d be willing to merge this? I could make the appropriate change to the convert-dictionaries.sh script of Plato (since it’s a relatively minor change). It would improve the speed of the process with very little work.