I like that README is added (I've had this chicken-and-egg problem with gemset.nix). However I think it is enough to do:

$ nix-shell -p bundix --run 'bundix'

to rebuild gemset.nix.

Also, why should we add rouge? There already is 'coderay', 'pygments.rb'. Which benefits has rouge compared to those?

I've tried pygments a while back with asciidoctor and it just didn't work.. Found this to be an easy workaround for what I needed instead of debugging pygments (or maybe failure to use it?)

On a more global scale, adding it here costs little and gives users more choice what to use

Regarding bundix, I've added nix-shell to be able to regenerate it without having the package installed.. As I stumbled on it a few times - otherwise it fails on missing dependencies.

As I stumbled on it a few times - otherwise it fails on missing dependencies.

I've tried rm gemset.nix && nix-shell --pure -p bundix --run 'bundix' and it was enough to regenerate gemset.nix. The rebuild-gemset.nix is redundant here, just mention the command I've proposed in README.adoc.

On a more global scale, adding it here costs little and gives users more choice what to use

Ok. I don't know Ruby packaging details, I'm curious how it behaves when I use asciidoctor and rouge in local Gemfile/gemset.nix. I hope there will be no conflicts.

Thanks, I'll try reproducing the dependencies error when regenerating. Will report back here when done. Will update readme too.

@svalaskevicius have you reproduced the issue?

Sorry didn't have a chance yet. Will try to do it this evening :)

ok, reproduced, (@danbst)

your command works ok only if Gemfile.lock is present, however, it doesn't add new libs to it, only converts the lock file to gemset.nix.

to add rouge to the lock file, one apparently needs to run bundix -m, which runs ruby things.. that fail to install mathematical because there are no binary dependencies installed:

<compiling things and dep errors>
An error occurred while installing mathematical (1.6.12), and Bundler cannot continue.
Make sure that `gem install mathematical -v '1.6.12' --source 'https://rubygems.org/'` succeeds before bundling.                                                             

Thus, I still don't see any other alternative rather than in this PR.

I'll update the readme to adoc though :)

done. also when testing now I noticed that rouge and some other dep have upped their version

@svalaskevicius thanks! Makes sense. I found a way to remove duplication. Here's diff:

diff --git a/pkgs/tools/typesetting/asciidoctor/default.nix b/pkgs/tools/typesetting/asciidoctor/default.nix
index 9508f3ff023..e46ebcf18ae 100644
--- a/pkgs/tools/typesetting/asciidoctor/default.nix
+++ b/pkgs/tools/typesetting/asciidoctor/default.nix
@@ -1,10 +1,10 @@
-{ stdenv, lib, bundlerApp, ruby
+{ stdenv, lib, bundlerApp, ruby, bundix, mkShell
   # Dependencies of the 'mathematical' package
 , cmake, bison, flex, glib, pkgconfig, cairo
 , pango, gdk_pixbuf, libxml2, python3, patchelf
 }:
 
-bundlerApp {
+bundlerApp rec {
   inherit ruby;
   pname = "asciidoctor";
   gemdir = ./.;
@@ -43,6 +43,12 @@ bundlerApp {
     };
   };
 
+  passthru.updateShell = mkShell {
+    buildInputs = (gemConfig.mathematical {}).buildInputs ++ [
+      bundix
+    ];
+  };
+
   meta = with lib; {
     description = "A faster Asciidoc processor written in Ruby";
     homepage = http://asciidoctor.org/;
diff --git a/pkgs/tools/typesetting/asciidoctor/update.sh b/pkgs/tools/typesetting/asciidoctor/update.sh
new file mode 100755
index 00000000000..10a053a847b
--- /dev/null
+++ b/pkgs/tools/typesetting/asciidoctor/update.sh
@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+rm gemset.nix Gemfile.lock
+nix-shell ../../../.. -A asciidoctor.updateShell --run '
+  bundix -m --bundle-pack-path $TMPDIR/asciidoctor-ruby-bundle
+'
+rm -r .bundle
\ No newline at end of file

That can also remove the need for README, as update.sh is kinda standard way for package update scripts.

If you like it, you may embed it into this PR. Otherwise I can push it separately.

Thanks! I'll test and add it a bit later :)

works well! I've set you as the author in the commit @danbst if it's ok with you :)

@svalaskevicius thanks for doublechecking!