Merge lp:~mbp/python-oops-datedir-repo/bsondump into lp:python-oops-datedir-repo

[Martin Pool]

This copies in the little bsondump script to make oops files readable.

It's a bit unclear which of the many oops components this really belongs in - it's not specific to storing it in per-date directories - but wgrant wanted it here, so here it is.

[Gavin Panella (allenap)]

Typically a script like this would go in utilities/ and the first thing it would do is import _pythonpath. Alternatively you could put it in and use buildout's templating to get the path setup. In either case you could then rely on having access to the same bson library that Launchpad depends on (bson 0.3.2 right now), and the code could cry a little less :)

You could consider using fileinput to deal with stdin and/or named files, though I'm not sure how well bson would cope with a stream containing multiple documents (for want of a better term).

Something like this might be a worthy addition to https://github.com/mongodb/mongo-python-driver (upstream for the python-bson package), so that we could use, say, `python -m bson/dump` anywhere python-bson is installed.

review: Needs Fixing

[Gavin Panella (allenap)]

> Alternatively you could put it in and use buildout's

s/it in and/it in buildout-templates and/

[Robert Collins (lifeless)]

Actually, we currently use the pypi bson package, not the mongo one, which is a wart itself. I don't want this stalled though, so I'm going to JFDI it as a regular buildout entry point.

[Robert Collins (lifeless)]

Other issues (fixing, but noting for easier ride next time:)

the new script was missing (C) header stuff (nab from any other file in the project).

No tests (I'm ignoring this for now given it's triviality... I bet this will bite later tho!)

bson doesn't handle multiple documents well, so using named files only is fine IMO.

review: Approve

[Martin Pool (mbp)]

thanks!

[Martin Pool (mbp)]

... as a follow on, we could delete it from lp:launchpad scripts/

[Robert Collins (lifeless)]

On Wed, Feb 1, 2012 at 7:38 PM, Martin Pool <email address hidden> wrote:
> ... as a follow on, we could delete it from lp:launchpad scripts/

please do!

[Martin Pool (mbp)]

https://code.launchpad.net/~mbp/launchpad/remove-bsondump2/+merge/93778

=== modified file '.bzrignore'
--- .bzrignore	2011-11-11 04:48:53 +0000
+++ .bzrignore	2011-12-20 05:30:30 +0000
@@ -9,3 +9,4 @@
 ./dist
 ./MANIFEST
 .testrepository
+./build

=== added directory 'scripts'
=== added file 'scripts/bsondump'
--- scripts/bsondump	1970-01-01 00:00:00 +0000
+++ scripts/bsondump	2011-12-20 05:30:30 +0000
@@ -0,0 +1,30 @@
+#! /usr/bin/python
+
+"""Print a BSON document for easier human inspection.
+
+This can be used for oopses, which are commonly (though not necessarily)
+stored as BSON.
+
+usage: bsondump FILE
+"""
+
+import bson
+from pprint import pprint
+import sys
+
+if len(sys.argv) != 2:
+    print __doc__
+    sys.exit(1)
+
+
+def decode_somehow(filename):
+    """There are several bson apis: try to decode it while crying."""
+    fn = getattr(bson, 'loads', None) or getattr(bson, 'decode_all')
+    return fn(file(filename, 'rb').read())
+
+
+# I'd like to use json here, but not everything serializable in bson is
+# easily representable in json - even before getting in to the weird parts,
+# oopses commonly have datetime objects. -- mbp 2011-12-20
+
+pprint(decode_somehow(sys.argv[1]))

=== modified file 'setup.py'
--- setup.py	2011-12-08 03:05:06 +0000
+++ setup.py	2011-12-20 05:30:30 +0000
@@ -30,6 +30,9 @@
       url="https://launchpad.net/python-oops-datedir-repo",
       packages=['oops_datedir_repo'],
       package_dir = {'':'.'},
+      scripts = [
+          'scripts/bsondump',
+          ],
       classifiers = [
           'Development Status :: 2 - Pre-Alpha',
           'Intended Audience :: Developers',