Commit by micah :: r11268 fidtool/:

Fidtool has been swallowed by CIA

Commit by micah :: r11266 fidtool/ (test.py timestamps.txt):

Slight tweak to the test dataset.

Commit by micah :: r9369 fidtool/_fidtool.c:

Fix a couple dangling references returned by PyIter_Next() during fid_graph_init()

Commit by micah :: r9299 fidtool/ (_fidtool.c test.py):

Added a new system for defining grids using any iterable object. The test.py example now uses a simple generator and the datetime module to produce a day-aligned grid in the local timezone

Commit by micah :: r9290 fidtool/ (_fidtool.c test.py):

Implement happy palette initialization with alpha blending and grid lines. This needs a more
accurate algorithm for placing the gridlines, but the C backend for fidtool is close to
feature-complete.

Commit by micah :: r9289 fidtool/ (_fidtool.c test.py):

Yay, graphs, with pretty antialiasing

Commit by micah :: r9288 fidtool/ (_fidtool.c test.py):

Add a thin buffering layer between libpng and the Python file-like object to avoid jillions of little 2 to 8 byte or so write()s. Now we just end up with one or more large writes for image data, one for headers, one for footers.

Commit by micah :: r9287 fidtool/ (_fidtool.c setup.py test.py):

Implemented the glue necessary to safely use libpng with python. Libpng errors turn into Python exceptions, libpng I/O is directed to any python file-like object, and exceptions from the file-like object are propagated out past libpng into the caller.

Commit by micah :: r9286 fidtool/_fidtool.c:

Add a header to the first L1 page, identifying the format version and page size. Return both sample and sample index when performing a set of seek queries.

Commit by micah :: r9282 fidtool/ (_fidtool.c test.py):

Reset the L0 watermark when we need it to re-seek, don't touch 'l0_cursor.sample'. This seems to be the last (major?) bug in the seek and append operations. test.py now runs smoothly on the whole CIA dataset.

Commit by micah :: r9281 fidtool/ (_fidtool.c test.py):

This is reeeeeally close to working at the L2 level, but it's time to take Chippy home, for real this time.

Commit by micah :: r9280 fidtool/ (_fidtool.c test.py):

Whee, L1 appends and seeks appear to work

Commit by micah :: r9279 fidtool/ (_fidtool.c test.py):

Yay, I finally understand how fidtool's L0 seek algorithm should work.
It looks like I've fixed all the ugly corner cases that were showing up
in test.py.

Commit by micah :: r9278 fidtool/_fidtool.c:

Almooost there... almost all the nasty fencepost bugs in L0 seek seem to be gone

Commit by micah :: r9277 fidtool/ (_fidtool.c test.py):

A little more cleanup and commenting on the L2 through L1 seek process

Commit by micah :: r9257 fidtool/_fidtool.c:

Wrote 120 lines of comments (with ascii art) explaining how the FID data structure should work. It does not yet work like this.

Commit by micah :: r9243 fidtool/ (_fidtool.c test.py):

Fixed some more bugs, but test.py still shows what looks like a sample index discontinuity as we cross L1 page boundaries. The root of this bug is probably the somewhat ambiguous way we define timestamp deltas on L1 pages. Fixing that is a job for tomorrow.

Commit by micah :: r9236 fidtool/ (_fidtool.c test.py):

Implement basic interval queries, fix a few subtle bugs in indexing. This now seems to mostly-work, but still has some subtle bugs that test.py is reporting.

Commit by micah :: r9226 fidtool/ (_fidtool.c test.py):

Appending across L0 page boundaries with updates to the L1 index now works. Can't yet append across L1 boundaries.

Commit by micah :: r9225 fidtool/ (_fidtool.c test.py):

This file format makes my head hurt. Appending should work, as long as you don't cross an L0 page boundary. Changed the semantics of the reverse-header on an L1 page just a bit, to make it quicker to encode