Handling jpx images with multiple precincts per resolution level

[plroit]

As discussed in the comment thread for issue 5475
introduced a small change at how codeblocks of each resolution are mapped to their precincts,
for the non-degenerate case.
Has been tested with large images (5K x 3K approx.) RGB, multiple quality layers, 6 resolution levels
with the following test cases:

1. No precincts at all (single default precinct for each resolution level)
2. Multiple precincts per resolution level with varying sizes for each resolution
3. Precincts with a size larger than the image (should be the similar to the degenerate case).

[Snuffleupagus]

Please remove the second commit, since it doesn't contain anything that's needed for this PR.
Also, please update the code so that it passes lint (run node make lint locally), see the Style Guide.
Once you've addressed these comments, please squash the commits: https://github.com/mozilla/pdf.js/wiki/Squashing-Commits.

[CodingFabian]

i can confirm that this commit makes https://www.dropbox.com/s/wnkhdtqgpduk5fl/cabs_1tile_bw_puniform_lrcp.pdf?dl=0 render correctly. well done!
We should add the pdf as regression test.

Besides that the linting errors are that your comment lines are too long (80 char limit) and we use === instead of ==

[Snuffleupagus]

We should add the pdf as regression test.

Ideally yes, but we need to obtain permission first. I've just asked about this in the issue.
(In this case a linked test is unfortunately not such a great idea, since the file is only available from a Dropbox which means a very uncertain future availability.)

[CodingFabian]

this looks good. can you please squash the commits? (https://github.com/mozilla/pdf.js/wiki/Squashing-Commits)

[yurydelendik]

nit: space after ':'

[yurydelendik]

Can you rename cblkGroupWidth to subbandCodeblockWidth and cblkGroupHeight to subbandCodeblockHeight?

Per "B.7 Division of the sub-bands into code-blocks": "The code-block size for each sub-band at a particular resolution level is determined as 2^xcb ́ by 2^ycb ́ ...."

[plroit]

Hi,
Section B. 6 specifies how to divide a resolution level into rectangular blocks called precincts, and section B.7 specifies how to divide a subband into rectangular blocks called codeblocks. Surprisingly, there is no direct specification of how to map codeblocks into precincts. That is why the Jasper library had to establish a new term, called _Codeblock-Group_ to derive the correct mapping, and I have decided to use this variable name.
For this reason I have to disagree with you about the naming, but I will change it if you should insist.

Please note that subbandCodeblockWidth/Height is a misleading variable name, since it does not describe neither the codeblock dimensions nor the subbands' dimensions.
What it should describe is the number of samples in a subband that map to a whole precinct.
Maybe a more descriptive name is subbandSamplesPerPrecinctWidth/Height ?

Paul

[yurydelendik]

Okay, subbandSamplesPerPrecinctWidth is good too (but takes more space and might add more code lines). If it's too ugly, will subbandWidthPerPrecinct work?

[yurydelendik]

Looks good after comments above addressed and commits squashed.

[plroit]

I am new to Git, so squashing the commits did not go very smoothly,
please make sure that the commit history looks like what you had in mind.

Thank you, Paul

[CodingFabian]

yeah looks quite a mess :-) the main issue is that you did the changes on master, rather an issue branch :)
I am going to save your work and make a new PR, but I think you can manage to recover from it. What I would do is: save your work, re-clone, make a branch and create a new pr.

[yurydelendik]

Fixed by #5485