kolibri
kolibri copied to clipboard
learningequality
PDF renderer shows wrong page size
Observed behavior
http://35.227.71.104/learn/#/topics/c/c1b42abb8ee45d2596e79197c46cd4c6
Expected behavior
The PDF file was tested in several other PDF readers and page always renders fine (half width as expected).
User-facing consequences
The PDFs of this specific channel will be difficult to read.
Steps to reproduce
Try viewing http://35.227.71.104/learn/#/topics/c/c1b42abb8ee45d2596e79197c46cd4c6 in full screen mode
Context
This PDF in the above links was obtained from a double-width PDF which was then cropped.
Here is a link to the original PDF http://www.pointb.is/s/21CSGuide_English.pdf and this is the code that cuts each page into left and right halves: https://github.com/learningequality/sushi-chef-pointb-21csguide/blob/master/sushichef.py#L28-L62
Tested on latest: release-v0.12.x and 0.11.1
The PDF renderer appears to be reserving 2x the space for the right-side pages (with binding on the left). The renderer must be inferring this page width from something in the PDF file.
Is there anything different about these right-side pages when they're being cheffed? It's as if they are still using the same dual-page-width dimensions from the original PDF.
@jonboiser Yes, the original PDF is 2x sized, so the chef code performs a "crop" operation by setting the cropBox and trimBox to half size, see
https://github.com/learningequality/sushi-chef-pointb-21csguide/blob/master/sushichef.py#L51-L54 (and mediaBox too but not show in this code)
The PDF rendering in any other PDF reader is fine (shows half-page), but somehow PDF.js doesn't seem to respect the page crop info, hence filing the issue here.
See #sushi-chefs threads:
- https://learningequality.slack.com/archives/C6S90FCDT/p1567402734002400
- https://learningequality.slack.com/archives/C6S90FCDT/p1567587911006900
We can also use the PDF.js demo site here to validate the .pdf rendering: https://mozilla.github.io/pdf.js/web/viewer.html - just open the .pdf file you want to test.
Thx @cpauya ; the above screenshot confirm that a more recent version of PDF.js would fix this issue.
The Kolibri document_pdf_render is pinned to 1.9.x:
https://github.com/learningequality/kolibri/blob/develop/kolibri/plugins/document_pdf_render/package.json#L7 which is 2 years old according to npm https://www.npmjs.com/package/pdfjs-dist
I will let the Kolibri team investigate further—there might be API differences or other obstacles to updated, but definitely worth looking into.
I tried to upgrade PDF.js at one point and everything broke. We should do it, but it will not be straight forward.
