Encoding issues with Hakyll & Nix

June 2020

When setting up this blog I already had a good experience with Hakyll and decided to use it again but now I wanted to try building it with Nix. Fortunately, there’s a great guide how to do that, except the compiled site generator won’t work with non-ASCII UTF-8 source files:

[ERROR] _cache/91cc3e6aa55338437b161dc62c21b154 for Hakyll.Core.Resource.Provider.MetadataCache/contact.markdown/body: Store.set: invalid argument (invalid byte sequence)

One option is to move out of the “small hut in the mountains of Kumano Kodō on Kii Hantō” (where you live according to Hakyll’s default contact.markdown) to a plain ASCII place. The other is to build the Hakyll site generator with support for more locales.

Being a total Nix rookie I mostly achieved that by consulting with nprindle on the Nix Discord server and by copying a few lines from other Nix scripts. The final result is to replace the hakyll-nix-example portion of release.nix with

hakyll-nix-example = (hpNew.callCabal2nix "hakyll-nix-example" ./. { }).overrideAttrs (old: {
      nativeBuildInputs = old.nativeBuildInputs or [] ++ [
      postInstall = old.postInstall or "" + ''
        wrapProgram $out/bin/site \
          --set LANG "en_US.UTF-8" \
          --set LOCALE_ARCHIVE "${pkgs.glibcLocales}/lib/locale/locale-archive"

nprindle was also building his blog at the time and you might find other neat tricks like proper cleaning of input to build (see clean) in his nix setup.