Log file Viewer
Help
|
Karma
| Search:
#pil logs for Tuesday the 30th of September, 2014
(
Back to #pil overview
) (
Back to channel listing
) (
Animate logs
)
[
05:38:43
] <
travis-ci
> [travis-ci] wiredfool/Pillow#507 (2.6.x - 1a91078 : Eric Soroos): The build passed.
[
05:38:43
] <
travis-ci
> [travis-ci] Change view : https://github.com/wiredfool/Pillow/compare/81ebc21abfdd...1a91078154a1
[
05:38:43
] <
travis-ci
> [travis-ci] Build details : http://travis-ci.org/wiredfool/Pillow/builds/36637940
[
10:09:50
] <
travis-ci
> [travis-ci] hugovk/Pillow#643 (master - 5a68276 : wiredfool): The build passed.
[
10:09:50
] <
travis-ci
> [travis-ci] Change view : https://github.com/hugovk/Pillow/compare/72c7073d31ba...5a6827690b7e
[
10:09:50
] <
travis-ci
> [travis-ci] Build details : http://travis-ci.org/hugovk/Pillow/builds/36653717
[
11:22:31
] <
travis-ci
> [travis-ci] hugovk/Pillow#645 (landscape - 1dfffd6 : Hugo): The build passed.
[
11:22:31
] <
travis-ci
> [travis-ci] Change view : https://github.com/hugovk/Pillow/compare/8804d778b4a9...68cf797c02f2
[
11:22:31
] <
travis-ci
> [travis-ci] Build details : http://travis-ci.org/hugovk/Pillow/builds/36659723
[
11:31:50
] <
travis-ci
> [travis-ci] hugovk/Pillow#646 (landscape - 4824bc6 : Hugo): The build passed.
[
11:31:50
] <
travis-ci
> [travis-ci] Change view : https://github.com/hugovk/Pillow/compare/104899f7aaa4...4824bc6e99b7
[
11:31:50
] <
travis-ci
> [travis-ci] Build details : http://travis-ci.org/hugovk/Pillow/builds/36660008
[
15:47:49
] <
travis-ci
> [travis-ci] wiredfool/Pillow#508 (2.6.x - 1bb8504 : wiredfool): The build passed.
[
15:47:49
] <
travis-ci
> [travis-ci] Change view : https://github.com/wiredfool/Pillow/compare/1a91078154a1...1bb850427dca
[
15:47:49
] <
travis-ci
> [travis-ci] Build details : http://travis-ci.org/wiredfool/Pillow/builds/36682628
[
16:44:55
] <
travis-ci
> [travis-ci] python-pillow/Pillow#1572 (2.6.x - 69994b1 : Hugo): The build passed.
[
16:44:55
] <
travis-ci
> [travis-ci] Change view : https://github.com/python-pillow/Pillow/compare/81ebc21abfdd...69994b11fb76
[
16:44:55
] <
travis-ci
> [travis-ci] Build details : http://travis-ci.org/python-pillow/Pillow/builds/36687925
[
19:02:16
] <
MathFox
> What is the deeper meaning behind the PIL.Image._MODEINFO fields; especially basetype...
[
19:20:42
] <
wiredfool
> basetype is the mode for an individual band in an image.
[
19:20:52
] <
wiredfool
> so, RGB is composed of 3 L images
[
19:21:07
] <
wiredfool
> RGB48 would be 3 * I;16
[
19:21:12
] <
wiredfool
> of one form or another
[
19:21:56
] <
MathFox
> But I;16 declares I as basetype...
[
19:22:18
] <
wiredfool
> ummm
[
19:22:42
] <
MathFox
> in ImageMode.py
[
19:23:39
] <
wiredfool
> yeah, that looks odd
[
19:25:08
] <
wiredfool
> ImageMode hasn't been (significantly) touched since the pillow fork
[
19:26:03
] <
wiredfool
> I'm going to go out on a limb and say that ImageMode is irrelevant, or close to it
[
19:26:26
] <
MathFox
> TIFF has orthogonality in "channel depth" and "channel interpretation"
[
19:27:51
] <
wiredfool
> provisionally, yes
[
19:28:00
] <
MathFox
> I think it is not very relevant to look for test cases that break I;16...
[
19:28:29
] <
wiredfool
> They show up from time to time
[
19:28:58
] <
wiredfool
> there's one open issue on a gis tiff format that I haven't had a chance to look at yet
[
19:29:33
] <
MathFox
> so that RGB and CYMK can be represented as bytes, floats or other types...
[
19:29:40
] <
wiredfool
> So, I don't see channel interpretation in the spec
[
19:29:47
] <
wiredfool
> nor in tiff image plugin.
[
19:30:29
] <
wiredfool
> ah, right.
[
19:30:40
] <
wiredfool
> Photometric Interpretation vs channel depth
[
19:30:55
] <
wiredfool
> yes, you are correct there
[
19:31:15
] <
wiredfool
> You've seen TiffImagePlugin.OPEN_INFO
[
19:32:09
] <
wiredfool
> you
[
19:32:17
] <
wiredfool
> yo're going to need an entry there like:
[
19:33:15
] <
wiredfool
> (II, 2, 1, 1, (16,16,16), ()): ('RGB;48','RGB:48
[
19:33:17
] <
wiredfool
> )
[
19:33:24
] <
MathFox
> (II, 2, 1, 1, (16, 16, 16), ()): ("RGB48", "RGB48"),
[
19:33:28
] <
wiredfool
> (roughly
[
19:34:01
] <
wiredfool
> right
[
19:34:31
] <
MathFox
> I could special code it through PIL, but I think it's wiser to spend a bit more time on the foundation
[
19:35:00
] <
wiredfool
> What I'd suggest is that we need one multichannel 16bit format working
[
19:35:18
] <
wiredfool
> as internal storage in Pillow, and likely interchange with num/scipy
[
19:35:19
] <
MathFox
> without too many special casing
[
19:36:06
] <
wiredfool
> Most of the different channel types wrt rgb/cmyk/hsv/lab are just the same infrastructure with different conversions
[
19:36:14
] <
wiredfool
> and different band names
[
19:36:17
] <
MathFox
> Make uint16 a full member of PIL
[
19:36:44
] <
wiredfool
> you mean not experimental?
[
19:38:07
] <
MathFox
> No, 16 bits/channel (or 16 bits/band) should be a base type
[
19:39:01
] <
MathFox
> It will be trivial to generalize to floats or 32 bits ints (/band) once the infrastructure is there
[
19:41:18
] <
wiredfool
> I'd have to think about how that would work
[
19:42:03
] <
wiredfool
> Currently, there are pretty much two core units, 8bit band and 32bit chunk.
[
19:42:26
] <
MathFox
> There is channels (R, G, B) or (L,) and underlying type L, I, F
[
19:42:59
] <
MathFox
> (Pity that L is overloaded to mean both "8 bits" and "Luminosity"
[
19:43:46
] <
wiredfool
> L is essentially I;8, but shorthand
[
19:44:16
] <
wiredfool
> (and L for LAB)
[
19:44:48
] <
wiredfool
> ((and when you get down to id, A is in LAB and Alpha))
[
19:50:12
] <
MathFox
> and whether a floating point red band needs "RF" or can just be "R" type "F"
[
19:50:41
] <
MathFox
> but then, an "F" band should be "L" type "F"
[
19:51:11
] <
wiredfool
> single band images are essentially luminance
[
19:51:21
] <
wiredfool
> unless you're doing something very odd
[
19:53:15
] <
MathFox
> but it only works for 8 bits depths :-(
[
19:54:00
] <
wiredfool
> Right.
[
19:54:28
] <
wiredfool
> one of the core we don't support multiband >8 bit depth things
[
19:54:52
] <
MathFox
> I'll ignore palleted images and bitmaps for now...
[
19:54:54
] <
wiredfool
> there's no storage, so nothing to split from, nor to merge too
[
19:55:20
] <
wiredfool
> lunchtime here. will return
[
19:55:38
] <
MathFox
> Add 16 and 32 bits RGB formats... and you'll break that barrier
[
20:04:06
] <
wiredfool
> yep
[
20:04:10
] <
wiredfool
> late lunch tho
[
20:04:23
] <
wiredfool
> had to chase a deer out of the garden earlier
[
20:09:00
] <
MathFox
> It's approaching bed time for me... West coast too :-)
[
20:10:16
] <
wiredfool
> maybe, but it's not deer season here, and he's eating my fruit trees