piston-mini-client

Merge lp://staging/~jml/piston-mini-client/split-request-base-1041825 into lp://staging/piston-mini-client

split-request-base-1041825
Merge into trunk

Proposed by Jonathan Lange on 2012-09-14

Status:

Merged

Approved by:

Anthony Lenton on 2012-09-24

Approved revision:

Merged at revision:

Proposed branch:

lp://staging/~jml/piston-mini-client/split-request-base-1041825

Merge into:

lp://staging/piston-mini-client

Diff against target:

778 lines (+444/-184)

4 files modified

piston_mini_client/__init__.py (+427/-182)
piston_mini_client/tests/test_failhandlers.py (+15/-0)
piston_mini_client/tests/test_log_to_file.py (+1/-1)
setup.py (+1/-1)

To merge this branch:

bzr merge lp://staging/~jml/piston-mini-client/split-request-base-1041825

Undecided

Triaged

Link a bug report

Reviewer	Review Type	Date Requested	Status
Anthony Lenton		2012-09-14	Approve on 2012-09-24
Review via email: mp+124464@code.staging.launchpad.net

Commit message

Add PistonRequester, enabling access to piston_mini_client functionality on a URL-by-URL basis

Description of the change

PistonAPI does two really useful things:
1. It makes it really easy to define a Python front-end for an HTTP API
2. It has a bunch of stuff to make doing HTTP requests easier

Sometimes, especially with callback services like pkgme-service, it would be really really useful to do 2. without doing 1. If you take a look at pkgme-service, you'll see code written by the original author (hello achuni!) that duplicates stuff going on in piston-mini-client.

This branch extracts that core, requesting logic into a base class and make PistonAPI subclass it. The base class is called PistonRequester (better names welcome).

Backwards compatibility has been preserved. Everything that subclasses PistonAPI today should still work after this change lands.

I've also changed setup.py to use setuptools, so I could run tox in a virtualenv.

Thanks,
jml

lp://staging/~jml/piston-mini-client/split-request-base-1041825 updated on 2012-09-20

69. By Jonathan Lange on 2012-09-20: Update docstring.
70. By Jonathan Lange on 2012-09-20: Scheme not needed in PistonRequester methods.
71. By Jonathan Lange on 2012-09-20: Give the full URL in errors, not just the service root.
72. By Jonathan Lange on 2012-09-20: Don't have special timeout handling logic. Just take parameters.
73. By Jonathan Lange on 2012-09-20: As few class variables as we can get away with.
74. By Jonathan Lange on 2012-09-20: Don't do special log_filename consulting in PistonRequester.
75. By Jonathan Lange on 2012-09-20: Use PistonRequester, rather than subclass it.
76. By Jonathan Lange on 2012-09-20: Make all the rest of the attributes private.
77. By Jonathan Lange on 2012-09-20: Issue deprecation warnings.

Revision history for this message

Jonathan Lange (jml) wrote on 2012-09-20:

On 2012-09-14::

16:48:19 <achuni> jml: I thought PistonAPI was going to use PistonRequester as a client in the end?
16:48:33 <achuni> (or did we agree to go the other way around in the end?)
16:48:59 <jml> achuni: you mean, use composition instead of inheritance?
16:49:04 <achuni> yep
16:49:51 <jml> achuni: I'd be very happy to do that. It'll mean potentially breaking backwards compat w/ existing PistonAPI users
16:50:21 <jml> achuni: or alternatively wrapping every existing method (e.g. _prepare_headers, _get_proxy_info, ...)
16:50:25 <achuni> jml: even if you leave the current _get / _post / ... methods there?
16:50:35 <achuni> ah, right
16:50:36 <achuni> hm
16:50:39 <jml> achuni: your call.
16:51:38 <achuni> jml: yep, we'd need to wrap all methods, I think... I'd say the wrappers can be deprecated immediately
16:52:23 <jml> achuni: want warnings for the deprecations, or are docstrings enough?
16:52:52 <achuni> jml: like PistonSerializable._as_serializable... I'm not sure that's the recommended way, but it's what was done last time
16:53:11 <jml> achuni: cool. I'll do that then.

I've now done this. I haven't cleared up the deprecation warnings in the tests.

lp://staging/~jml/piston-mini-client/split-request-base-1041825 updated on 2012-09-20

78. By Jonathan Lange on 2012-09-20: Rename to DEPRECATED_ATTRIBUTES
79. By Jonathan Lange on 2012-09-20: Update deprecation message.
80. By Jonathan Lange on 2012-09-20: Don't emit deprecation warnings on the public attributes
81. By Jonathan Lange on 2012-09-20: Fix a forwarding bug.