From 56892c177ef779c2939f9998f58ed8bb0a929e54 Mon Sep 17 00:00:00 2001
From: Grische <2787581+grische@users.noreply.github.com>
Date: Fri, 5 Jun 2026 14:25:04 +0200
Subject: [PATCH] python3Packages.unstructured: bundle NLTK data to fix
 import-time download

unstructured/nlp/tokenize.py downloads the `averaged_perceptron_tagger_eng`
and `punkt_tab` NLTK corpora at import time unless they are already present on
`nltk.data.path`. In offline or read-only/sandboxed environments (such as a
systemd service with a read-only filesystem) this fails with
`OSError: [Errno 30] Read-only file system: '/nltk_data'` as soon as a consumer
imports a partition module (e.g. `unstructured.partition.epub` via open-webui).

Bundle the two required corpora through `nltk-data` and register the directory
on `nltk.data.path` with a small postPatch, so importing unstructured works
without network access or a writable home. The directory is named `nltk_data`
because unstructured's resolver only uses such paths verbatim.

Also import `unstructured.nlp.tokenize` in pythonImportsCheck so the build
itself exercises the corpora lookup and guards against regressions.

Assisted-by: Claude Opus 4.8 (1M context)
---
 .../python-modules/unstructured/default.nix   | 29 +++++++++++++++++--
 1 file changed, 26 insertions(+), 3 deletions(-)

diff --git a/pkgs/development/python-modules/unstructured/default.nix b/pkgs/development/python-modules/unstructured/default.nix
index 058da8228165..e936e156b501 100644
--- a/pkgs/development/python-modules/unstructured/default.nix
+++ b/pkgs/development/python-modules/unstructured/default.nix
@@ -2,6 +2,7 @@
   lib,
   buildPythonPackage,
   fetchFromGitHub,
+  symlinkJoin,
 
   # build-system
   setuptools,
@@ -30,6 +31,7 @@
   joblib,
   # jsonpath-python,
   nltk,
+  nltk-data,
   olefile,
   orderly-set,
   python-dateutil,
@@ -118,6 +120,18 @@
 }:
 let
   version = "0.18.31";
+
+  # unstructured downloads these NLTK corpora at import time unless they are already on
+  # nltk.data.path, which fails in offline or read-only builds. Bundle them and register
+  # the directory in postPatch. It must be named "nltk_data": unstructured's resolver
+  # uses paths ending in "nltk_data" as-is and appends "/nltk_data" to any others.
+  nltkData = symlinkJoin {
+    name = "nltk_data";
+    paths = with nltk-data; [
+      averaged-perceptron-tagger-eng
+      punkt-tab
+    ];
+  };
 in
 buildPythonPackage rec {
   pname = "unstructured";
@@ -133,6 +147,11 @@ buildPythonPackage rec {
 
   build-system = [ setuptools ];
 
+  postPatch = ''
+    substituteInPlace unstructured/nlp/tokenize.py \
+      --replace-fail 'import nltk' 'import nltk; nltk.data.path.append("${nltkData}")'
+  '';
+
   dependencies = [
     # Base dependencies
     anyio
@@ -257,10 +276,14 @@ buildPythonPackage rec {
     ];
   };
 
-  pythonImportsCheck = [ "unstructured" ];
+  pythonImportsCheck = [
+    "unstructured"
+    # exercises the bundled NLTK corpora lookup, so the build catches an attempted download
+    "unstructured.nlp.tokenize"
+  ];
 
-  # test try to download punkt from nltk
-  # figure out how to make it available to enable the tests
+  # the import-time NLTK download is handled via nltkData above, but the test suite has
+  # further offline/data requirements that are not yet verified, so keep it disabled.
   doCheck = false;
 
   nativeCheckInputs = [