PDF viewer

Henning Hraban Ramm

22 Sep 2022 22 Sep '22

8:07 p.m.

Hi all, I compiled a kind of wishlist what we need or look for in a new or enhanced open source PDF viewer, as discussed e.g. at the 2021 meeting. https://wiki.contextgarden.net/PDF_viewer I’d like to further the subject to find a developer (or a team) and make this a reality. (Of course I’d be happy to have Michal on board!) Feel free to enhance the wiki page or discuss here. I suggest we’ll discuss funding in the context group board, but there was also a discussion about fundraising at the 2022 meeting that I didn’t follow – I think Paulo said he had experience in fundraising? Hraban

Show replies by date

Max Chernoff

23 Sep 23 Sep

2:47 a.m.

Hi Hraban,

...

I compiled a kind of wishlist what we need or look for in a new or enhanced open source PDF viewer, as discussed e.g. at the 2021 meeting.

https://wiki.contextgarden.net/PDF_viewer

Feel free to enhance the wiki page or discuss here.

I think that pdf.js (the Firefox PDF viewer) meets most of these requirements. There are a few requirements that it doesn't meet, but it's open source and written in JavaScript, so we should be able to add anything that's missing. Going down the list on the Wiki:

...

in general: - PDF 2.0 compatible

It can open PDF 2.0 files, but I doubt that it supports every single feature.

...

- proper color and transparency display (respect color profiles)

It supports all of the PDF colour models/profiles for vectors, although it ignores embedded colour profiles in images. https://github.com/mozilla/pdf.js/issues/2856 Regular RGB and CMYK stuff works fine though.

...

- limited JavaScript support (no modification, no file operations)

Yes, although it is quite limited: https://blog.mozilla.org/attack-and-defense/2021/10/14/implementing-form-fil...

...

- search

Yes.

...

- update on file change (configurable)

Not by default in Firefox. It works fine in this extension though: https://github.com/tomoki1207/vscode-pdfviewer

...

- keyboard control for as much as possible (but also menus/menubars)

Yes

...

- sidebar for bookmarks/ToC

Yes

...

- SyncTeX support

Not in Firefox, although there are patches to enable it: https://github.com/James-Yu/LaTeX-Workshop/blob/cff1a372/viewer/components/s... https://github.com/tomoki1207/vscode-pdfviewer/blob/92fecdb8/src/pdfPreview....

...

- accessibility (e.g. alternative text must work)

I've never used it, but it looks like there's good support: https://blog.mozilla.org/attack-and-defense/2021/10/14/implementing-form-fil...

...

- access to file attachments (lower priority)

Surprisingly, yes.

...

for presentations: - presentation mode (full screen, no visible controls, on one monitor/beamer of choice)

Yes.

...

- JS for controlling optional content groups (OCGs, Viewer Layers)

You can toggle all the layers in the side panel. I don't think that you can toggle via buttons or JS though.

...

- inner- and inter-document links must work

Yes; however, inter-document links are sometimes limited depending for security reasons. I'm sure that this can be configured though.

...

- multimedia support at least via links (suggestion: use “plain Rich Media” approach with OS’ default player)

I don't think so, although I guess you could include a video file as an attachment.

...

- transition effects not necessary (but if the developer has fun including them, go on)

No transition effects.

...

for forms: - fill in

Yes.

...

- print documents with/without form contents

Yes.

...

- same-ID fields on different pages must be synchronized

No idea.

...

- custom checkmarks/radiobuttons should work & display correctly

Usually it works, sometimes it doesn't.

...

- JS for calculations

Usually it works, sometimes it doesn't.

...

for annotations (correction workflow; generally just nice to have): - similar to Adobe/Foxit Reader

Reading annotations works, but you can't modify anything.

...

for development/debugging: - show metadata (incl. XMP), e.g. in a sidebar or dialog

You can see all the "regular" metadata, but XMP doesn't work.

...

- show fonts metadata (inclusion etc.)

No.

...

- show PDF Boxes (MediaBox, CropBox, TrimBox, BleedBox, ArtBox) as colored frames, switchable via preferences and menu (bar) item

No.

...

- check for PDF/A and PDF/X compliance would be nice (integrate VeraPDF?)

...

- FOSS license

Yes. (Apache 2.0)

...

- should at least work on Windows, MacOS (preferably 10.14+), Linux and *BSD; iOS & Android versions would be nice

It's built in to Firefox, so it works on Windows, macOS, Linux, and the BSDs. It should also work on mobile, although support is a little spottier there.

...

- how about (long term) support?

It should be supported as long as Firefox exists.

...

- Funding

Overleaf uses pdf.js as its default PDF viewer, so they may perhaps be willing to help. ~~~~~~~~~~ I've been using pdf.js almost exclusively for the past few years either via Firefox or VS Code, and I've never really had any problems. The only real issue that I've had is that it gets fairly slow with documents over a few thousand pages long. Otherwise, it seems pretty fast and stable, and it supports nearly every feature that I tend to need. Since it's written in JavaScript, it should be fairly easy to modify, and it should run on nearly everything. -- Max

Alan Braslau

4:30 a.m.

On Thu, 22 Sep 2022 18:47:50 -0600 Max Chernoff via ntg-context wrote:

...

I've been using pdf.js almost exclusively for the past few years either via Firefox or VS Code, and I've never really had any problems. The only real issue that I've had is that it gets fairly slow with documents over a few thousand pages long. Otherwise, it seems pretty fast and stable, and it supports nearly every feature that I tend to need.

I have never used firefox to display pdf files. I just tried it and find that it has an interesting feature of caching rendered pages. Pages with complicated graphics (think MetaPost with hundreds of thousands of objects) get rendered pretty quickly, although one can see the very many objects being drawn the first time the page is displayed. Subsequently displays of these pages have no lag whatsoever. Alan

Henning Hraban Ramm

9:59 a.m.

Am 23.09.22 um 02:47 schrieb Max Chernoff:

...

...
I compiled a kind of wishlist what we need or look for in a new or enhanced open source PDF viewer, as discussed e.g. at the 2021 meeting.

https://wiki.contextgarden.net/PDF_viewer

Feel free to enhance the wiki page or discuss here.

I think that pdf.js (the Firefox PDF viewer) meets most of these requirements. There are a few requirements that it doesn't meet, but it's open source and written in JavaScript, so we should be able to add anything that's missing.

You’re right, pdf.js is quite good, and it’s already used by a lot of projects (browsers, editor plugins etc.) – the latter also means that you often encounter old/buggy/limited versions. Unfortunately, Firefox doesn’t register itself as a PDF viewer (at least on MacOS), that means I can’t use it easily to open a PDF from the command line (e.g. in scripts). But anyway, that could shift our focus to add/enable features for pdf.js in TeX editors.

...

...
- PDF 2.0 compatible

It can open PDF 2.0 files, but I doubt that it supports every single feature.

I won’t expect every feature, there are too many. Actually I don’t know which features of PDF 2.0 Hans was missing in viewers.

...

...
- proper color and transparency display (respect color profiles)

It supports all of the PDF colour models/profiles for vectors, although it ignores embedded colour profiles in images.

https://github.com/mozilla/pdf.js/issues/2856

Regular RGB and CMYK stuff works fine though.

That’s a problem if you prepare PDFs for printing. Apple’s PDF library (used e.g. in Preview.app, Skim, TeXshop) also does a bad job at it, as well as Sumatra and Chrome (according to the comments to that ticket).

...

...
- limited JavaScript support (no modification, no file operations)

Yes, although it is quite limited:

https://blog.mozilla.org/attack-and-defense/2021/10/14/implementing-form-fil...

Thank you for the link! What they write about the scope of JS in pdf.js sounds good; but presentation stuff like buttons (e.g. for OCG) is probably missing. I need to check the limits of forms in pdf.js; but Mozilla even implemented XFA (e.g. Qoppa didn’t).

...

...
- update on file change (configurable)

Not by default in Firefox. It works fine in this extension though:

https://github.com/tomoki1207/vscode-pdfviewer

There it’s probably the editor framework that enables refresh. (Same experience with pdf.js in Atom.)

...

...
- SyncTeX support

Not in Firefox, although there are patches to enable it:

https://github.com/James-Yu/LaTeX-Workshop/blob/cff1a372/viewer/components/s... https://github.com/tomoki1207/vscode-pdfviewer/blob/92fecdb8/src/pdfPreview....

One of the Atom plugins was supposed to have it, but it didn’t work... (SyncTeX with ConTeXt is special anyway, we need to look into that again.)

...

...
- accessibility (e.g. alternative text must work)

I've never used it, but it looks like there's good support:

https://blog.mozilla.org/attack-and-defense/2021/10/14/implementing-form-fil...

Sounds good.

...

...
- JS for controlling optional content groups (OCGs, Viewer Layers)

You can toggle all the layers in the side panel. I don't think that you can toggle via buttons or JS though.

Same as in Okular. OCG control via JS would enable stepwise presentations and simple animations.

...

...
- inner- and inter-document links must work

Yes; however, inter-document links are sometimes limited depending for security reasons. I'm sure that this can be configured though.

Of course; I just meant it would be handy to be able to jump between PDFs in the same directory.

...

...
- multimedia support at least via links (suggestion: use “plain Rich Media” approach with OS’ default player)

I don't think so, although I guess you could include a video file as an attachment.

A question for Michal...

...

...
- transition effects not necessary (but if the developer has fun including them, go on)

No transition effects.

...
for forms: - fill in Yes.

Just checked again with current Firefox: It doesn’t work with all of my test files.

...

...
- print documents with/without form contents Yes.

I couldn’t find an option to print without form contents. (But usually you would want filled forms, so “with” is ok.)

...

...
- same-ID fields on different pages must be synchronized

No idea.

If the form generally works, this does also. I listed it because it’s a problem in some viewers.

...

...
- custom checkmarks/radiobuttons should work & display correctly

Usually it works, sometimes it doesn't.

...
- JS for calculations

Usually it works, sometimes it doesn't.

Need to check further...

...

...
for annotations (correction workflow; generally just nice to have): - similar to Adobe/Foxit Reader

Reading annotations works, but you can't modify anything.

Ok. There’s still no PDF viewer on Linux that can handle annotations well. (But even Acrobat Reader on MacOS frequently crashes on them; I’m using Foxit Reader for annotations, but the one for Linux is too old.)

...

...
for development/debugging: - show metadata (incl. XMP), e.g. in a sidebar or dialog

You can see all the "regular" metadata, but XMP doesn't work.

That should be easy to add, since pdf.js can handle Tagged PDF (XML parser included).

...

...
- show fonts metadata (inclusion etc.) No.

Pity. Should also not be too difficult, since the viewer needs the data anyway.

...

...
- show PDF Boxes (MediaBox, CropBox, TrimBox, BleedBox, ArtBox) as colored frames, switchable via preferences and menu (bar) item No.

Of course. There’s no viewer except Adobe Reader that shows these. But it’s an important feature for print preparation and should be easy to implement.

...

...
- Funding Overleaf uses pdf.js as its default PDF viewer, so they may perhaps be willing to help.

At least with SyncTeX support and other closely TeX-related stuff.

...

I've been using pdf.js almost exclusively for the past few years either via Firefox or VS Code, and I've never really had any problems. The only real issue that I've had is that it gets fairly slow with documents over a few thousand pages long. Otherwise, it seems pretty fast and stable, and it supports nearly every feature that I tend to need.

Well, documents with thousands of pages are probably unreliable/slow in most viewers.

...

Since it's written in JavaScript, it should be fairly easy to modify, and it should run on nearly everything.

Thank you! Hraban

Max Chernoff

12:18 p.m.

Hi Hraban,

...

Unfortunately, Firefox doesn’t register itself as a PDF viewer (at least on MacOS), that means I can’t use it easily to open a PDF from the command line (e.g. in scripts).

That's odd. You can set it as the default PDF viewer on Windows and Linux at least.

...

...
...
for forms: - fill in Yes.

Just checked again with current Firefox: It doesn’t work with all of my test files.

I tested it with the eforms manual: http://mirrors.ctan.org/macros/latex/contrib/acrotex/doc/eformman.pdf and it seems to work fine for me. I've also used it to fill out a few government forms in the past and it's worked too. Unfortunately, I think that there are like 12 incompatible ways of making a form in PDF, so support probably heavily depends on how the document was made.

...

...
...
- print documents with/without form contents Yes.

I couldn’t find an option to print without form contents. (But usually you would want filled forms, so “with” is ok.)

Well if you refresh the page, you can delete everything that you've filled in :)

...

...
...
- custom checkmarks/radiobuttons should work & display correctly

Usually it works, sometimes it doesn't.

...
- JS for calculations

Usually it works, sometimes it doesn't.

Need to check further...

I checked with the eforms manual linked above. Check marks and radio buttons seem to work, but calculations don't.

...

...
...
for annotations (correction workflow; generally just nice to have): - similar to Adobe/Foxit Reader

Reading annotations works, but you can't modify anything.

Ok. There’s still no PDF viewer on Linux that can handle annotations well. (But even Acrobat Reader on MacOS frequently crashes on them; I’m using Foxit Reader for annotations, but the one for Linux is too old.)

Microsoft Edge has decent PDF annotation support. I've never tested it on Linux, but a Linux version does exist. Okular also lets you add some annotations.

...

...
I've been using pdf.js almost exclusively for the past few years either via Firefox or VS Code, and I've never really had any problems. The only real issue that I've had is that it gets fairly slow with documents over a few thousand pages long. Otherwise, it seems pretty fast and stable, and it supports nearly every feature that I tend to need.

Well, documents with thousands of pages are probably unreliable/slow in most viewers.

I've got a 1.1GB document with 16000 pages, and Okular handles it just as fast as a 10 page document. Firefox at least manages to not crash when opening the document, which is better than most viewers. Okular is actually a pretty nice viewer in general. It's really fast, and it also seems to support most of these features. The tricky thing with it though is that I think that it would be much harder to modify compared to pdf.js. -- Max

Pablo Rodriguez

5:56 p.m.

On 9/23/22 12:18, Max Chernoff via ntg-context wrote:

...

...
Unfortunately, Firefox doesn’t register itself as a PDF viewer (at least on MacOS), that means I can’t use it easily to open a PDF from the command line (e.g. in scripts).

That's odd. You can set it as the default PDF viewer on Windows and Linux at least.

Hi Hraban, I know you wrote scripts, but here you have two links on how to set another PDF default viewer: https://discussions.apple.com/thread/7883441 https://www.igeeksblog.com/how-to-change-default-pdf-viewer-on-mac/ And to open PDF documents from scripts, once the default viewer is set, it seems that "open" is the right way to go (similar to "xdge-open" in Linux).

...

...
...
[…]

...
- JS for calculations

I would love that JS worked fine in PDF.js. I was surprised that not even this worked fine (https://github.com/mozilla/pdf.js/issues/13918): this.pageNum = this.numPages ; I would love that the kind of presentation contained in https://free-culture.tk could be displayed as https://www.vector-video.tk/bga-presvoz.pdf (only Acrobat seems to be able to deal with it). If PDF.js could deal with multimedia and JS in PDF, other audio formats than MP3 (such as Opus) could be used.

...

...
...
...
for annotations […] Reading annotations works, but you can't modify anything. […] Microsoft Edge has decent PDF annotation support. I've never tested it on Linux, but a Linux version does exist. Okular also lets you add some annotations.

Annotations may be part of a PDF viewer, but the problem is that in some projects they understand that the feature to add them is part of a PDF editor (so annotations won’t be added). I think this is what happened to annotations in Evince (but I may be wrong here). And I wonder whether PDF.js might take this approach. Just in case it might help, Pablo

Bruce Horrocks

5:34 p.m.

On 23 Sep 2022, at 08:59, Henning Hraban Ramm via ntg-context wrote:

...

Unfortunately, Firefox doesn’t register itself as a PDF viewer (at least on MacOS), that means I can’t use it easily to open a PDF from the command line (e.g. in scripts).

The following works for me in Monterey on an Intel Mac: $ /Applications/Firefox.app/Contents/MacOS/firefox test.pdf — Bruce Horrocks Hampshire, UK

juh

25 Sep 25 Sep

5:24 p.m.

Am Thu, Sep 22, 2022 at 06:47:50PM -0600 schrieb Max Chernoff via ntg-context:

...

I think that pdf.js (the Firefox PDF viewer) meets most of these requirements. There are a few requirements that it doesn't meet, but it's open source and written in JavaScript, so we should be able to add anything that's missing.

I found a nasty bug in firefox. https://mailman.ntg.nl/pipermail/ntg-context/2022/105212.html https://bugzilla.mozilla.org/show_bug.cgi?id=1781022 juh

Pablo Rodriguez

6:41 p.m.

On 9/25/22 17:24, juh via ntg-context wrote:

...

Am Thu, Sep 22, 2022 at 06:47:50PM -0600 schrieb Max Chernoff via ntg-context:

...
I think that pdf.js (the Firefox PDF viewer) meets most of these requirements. There are a few requirements that it doesn't meet, but it's open source and written in JavaScript, so we should be able to add anything that's missing.

I found a nasty bug in firefox.

https://mailman.ntg.nl/pipermail/ntg-context/2022/105212.html

https://bugzilla.mozilla.org/show_bug.cgi?id=1781022

As Jonathan Kew explained, TeX Gyre Heros is also poorly displayed with Acrobat. The issue may be trickier than we thought (since fixing Firefox with Heros+LMTX is only part of the issue). Just as a personal note, I tend to avoid using Heros with LMTX, since resulting PDF documents may be displayed (poorly on screen) with Acrobat. I wonder whether this is related to common code for both Acrobat and Firefox (originally comming from https://lists.nongnu.org/archive/html/freetype-devel/2013-05/msg00000.html). Pablo

luigi scarso

23 Sep 23 Sep

11:52 a.m.

On Thu, Sep 22, 2022 at 8:08 PM Henning Hraban Ramm via ntg-context < ntg-context@ntg.nl> wrote:

...

Hi all, I compiled a kind of wishlist what we need or look for in a new or enhanced open source PDF viewer, as discussed e.g. at the 2021 meeting.

https://wiki.contextgarden.net/PDF_viewer

slightly off-topic 1) gs has a new pdf viewer https://www.ghostscript.com/blog/pdfi.html but of course I think that they push on mupdf 2) EBookDroid - PDF & DJVU Reader is a very nice pdf viewer for android, especially on 10" tablets. It's based on mupdf. It seems that mupdf is more suited to be embedded into / used for embedding other programs: latest mupdf (1.20, iirc) has a js console and I think you can interact with the viewer -- sort of the js console of the browser; I haven't seen how yet, perhaps you can add objects on the fly to the documents, in this case one can think of embedding a small luatex (no harms, after all mupdf it's already a big binary) and add (pdf)objects by scripting. -- luigi

Pablo Rodriguez

5:23 p.m.

On 9/22/22 20:07, Henning Hraban Ramm via ntg-context wrote:

...

[…] I suggest we’ll discuss funding in the context group board, but there was also a discussion about fundraising at the 2022 meeting that I didn’t follow – I think Paulo said he had experience in fundraising?

Hi Hraban, sorry if I’m misreading, but I’m afraid I wasn’t able to find any Paulo in the messages I have from this list. Now I wonder who he might be. Many thanks for your help, Pablo

Willi Egger

26 Sep 26 Sep

4:44 p.m.

Hi Pablo, Hraban refers to Paulo Ney de Souza from US. He recently joined the ConTeXt meeting and indeed he mentioned that TUG had a fund raising campaign which was successful. Kind regards Willi

...

On 23 Sep 2022, at 17:23, Pablo Rodriguez via ntg-context wrote:

On 9/22/22 20:07, Henning Hraban Ramm via ntg-context wrote:

...
[…] I suggest we’ll discuss funding in the context group board, but there was also a discussion about fundraising at the 2022 meeting that I didn’t follow – I think Paulo said he had experience in fundraising?

Hi Hraban,

sorry if I’m misreading, but I’m afraid I wasn’t able to find any Paulo in the messages I have from this list.

Now I wonder who he might be.

Many thanks for your help,

Pablo ___________________________________________________________________________________ If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://www.ntg.nl/mailman/listinfo/ntg-context webpage : https://www.pragma-ade.nl / http://context.aanhet.net archive : https://bitbucket.org/phg/context-mirror/commits/ wiki : https://contextgarden.net ___________________________________________________________________________________

Pablo Rodriguez

7:15 p.m.

On 9/26/22 16:44, Willi Egger via ntg-context wrote:

...

Hi Pablo,

Hraban refers to Paulo Ney de Souza from US. He recently joined the ConTeXt meeting and indeed he mentioned that TUG had a fund raising campaign which was successful. Many thanks for your replies, Hraban and Willi.

Pablo

818

Age (days ago)

822

Last active (days ago)

List overview

Download

12 comments

8 participants

participants (8)

Alan Braslau
Bruce Horrocks
Henning Hraban Ramm
juh
luigi scarso
Max Chernoff
Pablo Rodriguez
Willi Egger

PDF viewer

tags

participants (8)