From: W. Trevor King <wking@drexel.edu>
Date: Fri, 17 Feb 2012 02:49:31 +0000 (-0500)
Subject: Add Unicode long_description in setup.py post.
X-Git-Url: http://git.tremily.us/?a=commitdiff_plain;h=50094f14c6f7df96603971828d47727818b74fd7;p=blog.git

Add Unicode long_description in setup.py post.
---

diff --git a/posts/Unicode_long_description_in_setup.py.mdwn b/posts/Unicode_long_description_in_setup.py.mdwn
new file mode 100644
index 0000000..b05ca4c
--- /dev/null
+++ b/posts/Unicode_long_description_in_setup.py.mdwn
@@ -0,0 +1,35 @@
+[[!meta  title="Unicode `long_description` in `setup.py`"]]
+
+I've been trying to figure out how to setup Unicode long descriptions
+in `setup.py`.  I often use Unicode in my `README` files and then use
+the contents of the `README` to set the long description with
+something like:
+
+    …
+    _this_dir = os.path.dirname(__file__)
+    …
+    setup(
+        …
+        long_description=codecs.open(
+            os.path.join(_this_dir, 'README'), 'r', encoding='utf-8').read(),
+        )
+
+This crashed in Python 2.7 with a `UnicodeDecodeError` when I tried to
+register the package on PyPI.  The problem is that packages are
+checked before registration to avoid being registered with broken
+metadata, and Unicode handling was broken in distutils ([bug
+13114][bug13114]).  Unfortunately, there haven't yet been Python
+releases containing the fixes (applied in October 2011).
+
+How do you work around this issue until get a more recent Python 2.7?
+Just use Python 3.x, where Unicode handling is much cleaner.  You may
+need hide Python-3-incompatible code inside:
+
+    if _sys.version_info < (3,0):
+
+blocks, but you shouldn't be pulling in huge amounts of code for
+`setup.py` anyway.
+
+[bug13114]: http://bugs.python.org/issue13114
+
+[[!tag tags/python]]