Support for older glibc

[dvwright]

Hi, first off thank you for this library, it's the only non-java languagetool alternative I've found.

Unfortunately, I am receiving an error when trying to use it,ImportError: /lib/x86_64-linux-gnu/libm.so.6: version 'GLIBC_2.27' not found (required by python/lib/python3.8/site-packages/nlprule.cpython-38-x86_64-linux-gnu.so)

I'm on a hosting environment where I don't have access to upgrade system libraries so i can't just upgrade glibc. The current version is glibc 2.19.

Is glibc 2.27 a hard requirement or is there a way to specify an older version of glibc?

I have a feeling this is Rust specific issue but I am new to Rust and not familiar with it's environment.

Thanks

[bminixhofer]

Hi, thanks.

Everything from glibc 2.11 can work: https://github.com/PyO3/maturin#manylinux-and-auditwheel. It seems I'll have to build the wheels in a Docker container in GH actions for full manylinux compliance. That shouldn't be too hard, I'll take a look at it.

[bminixhofer]

Hi, can you please try this wheel:

nlprule-0.6.1_pre-cp38-cp38-manylinux2014_x86_64.zip

(rename .zip to .whl, otherwise Github doesn't let me upload it)

This should work for glibc 2.17 and up. You can just try running the code in the README, if it works, I'll do a release.

[dvwright]

Hi @bminixhofer that wheel installed! Thanks

Unfortunately, I am receiving an exception on load, tokenizer = Tokenizer.load("en") ; OSError: failed to fill whole buffer.

This doesn't seem related to glibc, so if you like I can close this case as fixed and and open a new one for the buffer issue?

[bminixhofer]

Great. That's actually not an issue. There are no binaries uploaded for the prerelease so .load can't work. But the error message wasn't propagated properly. I fixed that now, if you install this wheel:

nlprule-0.6.1_pre-cp38-cp38-manylinux2014_x86_64.zip

You should see something like ValueError: HTTP status client error (404 Not Found) for url (https://github.com/bminixhofer/nlprule/releases/download/0.6.1-pre/en_tokenizer.bin.gz) when running Tokenizer.load("en").

Instead, you'd have to manually download the binaries from https://github.com/bminixhofer/nlprule/releases/tag/0.6.0, unzip and load with tokenizer = Tokenizer("/path/to/en_tokenizer.bin"), same for the Rules. Loading the wheel works so I'm pretty sure this should work too in your setup. But would still be great if you could confirm it works.

[dvwright]

Hi @bminixhofer I installed the new wheel and am getting the error you stated.

I am also getting the error if I specify a local path to the binaries I downloaded.

ValueError: HTTP status client error (404 Not Found) for url (https://github.com/bminixhofer/nlprule/releases/download/0.6.1-pre/en_rules.bin_tokenizer.bin.gz)
    tokenizer = Tokenizer.load("en_tokenizer.bin")

[dvwright]

It looks like if I specify a local path, it appends it to the home repo url, rather then using a local resource,file://,

ValueError: HTTP status client error (404 Not Found) for url (https://github.com/bminixhofer/nlprule/releases/download/0.6.1-pre//tmp/en_tokenizer.bin_tokenizer.bin.gz)
--
tokenizer = Tokenizer.load("/tmp/en_tokenizer.bin")

[bminixhofer]

Hi, you have to use Tokenizer(...) not Tokenizer.load(...) to load it from a local file, .load just takes a language code. Sorry, there should be a better error message here as well.

[dvwright]

Confirmed. This is working for me, thanks!