Support SVG images in multiple formats

[Nutomic]

I'm testing pandoc's markdown to pdf conversion using this readme and command ~/.cabal/bin/pandoc README.md -o test.pdf. However, I'm getting this error:

pandoc: Unable to convert image `/tmp/tex2pdf.26140/0056733ea8a4c9b7dab70a5b455f59437d34f134.svgz':
Cannot load file
Jpeg Invalid marker used. Failed reading at byte position 2
PNG Invalid PNG file, signature broken. Failed reading at byte position 8
Bitmap Invalid Bitmap magic identifier. Failed reading at byte position 2
GIF Invalid Gif signature : <svg x. Failed reading at byte position 6
HDR Invalid radiance file signature. Failed reading at byte position 11
Tiff Invalid endian tag value. Failed reading at byte position 2

pandoc: Unable to convert image `/tmp/tex2pdf.26140/8ba8f2c25e51f37aa52f6286040bc2a2ca70cda4.svgz':
Cannot load file
Jpeg Invalid marker used. Failed reading at byte position 2
PNG Invalid PNG file, signature broken. Failed reading at byte position 8
Bitmap Invalid Bitmap magic identifier. Failed reading at byte position 2
GIF Invalid Gif signature : <svg v. Failed reading at byte position 6
HDR Invalid radiance file signature. Failed reading at byte position 11
Tiff Invalid endian tag value. Failed reading at byte position 2

! Font T1/cmr/m/n/10=ecrm1000 at 10.0pt not loadable: Metric (TFM) file not fou
nd.
<to be read again> 
                   relax 
l.100 \fontencoding\encodingdefault\selectfont

pandoc: Error producing PDF from TeX source

Additionally, there's a missfont.log generated:

mktextfm ecrm1000
mktextfm ecrm1000

At first I thought this was because of the relative links for images, but replacing them with raw.githubusercontent.com URLs gives the same error.

Pandoc version is 1.3.11, installed from cabal, on Kubuntu 14.10.

[mpickering]

I can't reproduce this exact error, this is my output.

pandoc: Unable to convert image `/var/folders/3z/_vqy7kmx4pd90sg_v80zpk340000gn/T/tex2pdf.70131/0056733ea8a4c9b7dab70a5b455f59437d34f134.svgz':
Cannot load file
Jpeg Invalid marker used
PNG Invalid PNG file, signature broken
Bitmap Invalid Bitmap magic identifier
GIF Invalid Gif signature : <svg x
HDR Invalid radiance file signature
Tiff Invalid endian tag value

pandoc: Unable to convert image `/var/folders/3z/_vqy7kmx4pd90sg_v80zpk340000gn/T/tex2pdf.70131/8ba8f2c25e51f37aa52f6286040bc2a2ca70cda4.svgz':
Cannot load file
Jpeg Invalid marker used
PNG Invalid PNG file, signature broken
Bitmap Invalid Bitmap magic identifier
GIF Invalid Gif signature : <svg v
HDR Invalid radiance file signature
Tiff Invalid endian tag value

! LaTeX Error: Unknown graphics extension: .svgz.

See the LaTeX manual or LaTeX Companion for explanation.
Type  H <return>  for immediate help.
 ...                                              

l.63 ...33ea8a4c9b7dab70a5b455f59437d34f134.svgz}}

pandoc: Error producing PDF from TeX source

[Nutomic]

Yeah apparently the font error is a seperate problem (I'm also getting it for .md files without images).

[mpickering]

Can you compile normal tex documents?

[Nutomic]

It works using pdflatex in the shell. Although I don't know what pandoc uses.

[mpickering]

Pandoc uses pdflatex with the following arguments from what I can see "-halt-on-error", "-interaction", "nonstopmode","-output-directory"

[iliakan]

Same with epub/html to pdf.
Apparently, the problem is with SVG images <img src="image.svg"> which I use in the text.

[iliakan]

P.S. Just to note: there's no svgz in my HTML page that I try to convert, only svg, which is sent compressed-on-the-fly by the server though, headers:

Content-Encoding:gzip
Content-Type:image/svg+xml

[mpickering]

Can you please post the exact output you are getting in a new ticket?

[jgm]

I don't think svg is supported by pdflatex. The svg (or svgz) would need to be converted to a supported image format. We use JuicyPixels for bitmap image format conventions, but it won't help with svg.

I've improved the confusing error messages in a recent commit, but I don't think there's much else that can be done now.

Your path forward is to convert the svgs to pdfs with an external program.

[Twinside]

As a matter of fact, there is a possibility to render SVG, not directly through JuicyPixels, but I have a library called rasterific-svg for this. I was thinking of putting together a proposition to allow the use of SVG in pandoc, but I don't really know where to start.

[lierdakil]

@Twinside, rasterizing svg for pdf output seems somewhat suboptimal, when it could be converted straight to pdf (both being vector formats and all). But I don't think there are libraries for that...

[Twinside]

I think it's doable, I'll have a shot at it.

[Twinside]

rasterific-svg 0.2.3 now support PDF output (for a better function documentation see renderSvgDocument, it uses the same parameters).

[tarleb]

@Twinside Do you think SVG support could be added to pandoc using your library?

[Twinside]

There is #2211 proposed a long time ago, providing svg rasterization or svg conversion to PDF depending on the need. Conversion to PDF didn't handle inclusion of images though.

[tarleb]

I'm sorry, I hadn't noticed. Thanks for the pointer. Looks like the main obstacle is the indirect dependency on lens. @jgm and @mpickering both have raised their concern on the size of the lens library, maybe this discussion can be revisited to a definite conclusion?

From a quick look at the PR, it looks to me like gif and png would no longer be supported. Did I misread the change? I'll try to spend more time on this, it would be nice if we could get the issue at hand resolved.

[Twinside]

PNG & gif are still supported, it's the way to determine their DPI that would change, by using JuicyPixels's metadata information. But maybe I missed something (it was a loooong time ago).

[tarleb]

Figured I'd be missing something there, thanks.

it was a loooong time ago

😔

[JamesH65]

+1 for SVG conversion. Having real trouble finding a way of converting github markdown to PDF because of the lack of SVG support in pandoc.

[jgm]

Putting this in the pandoc 2.0 milestone, so I don't forget about it.
Not promising to implement it (the lens issue is still one to consider), but at least a decision should be made.

[mb21]

This would also be solved by switching the default PDF engine to ConTeXt or wkhtmltopdf...

@JamesH65 you can already use pandoc -t html5 -o foo.pdf (to use wkhtmltopdf) or pandoc -t context -o foo.pdf (to use ConTeXt) for PDF generation now.

[jgm]

This would also be solved by switching the default PDF engine to ConTeXt or wkhtmltopdf...

I don't think I'd want to change the default. But it's good to emphasize that people who use SVGs can already use `-t context` or `-t html` with `-o output.pdf` and use these engines, which support SVG already. I should mention another workaround, for those who need to support SVGs. Just use a program of your choice to convert the SVG to PDF, so you have two images, one with an svg extension, one with a pdf extension. Then leave off the extensions in your Markdown source and use `--default-image-extension` as appropriate for the output format. A filter would be another approach.

[ickc]

Another approach is to rely on an external dependencies, if including a big library is a concern. And since LaTeX is already an external dependency of pandoc, I guess this will be fine for many potential users of this feature.

e.g. in graphics - How to include SVG diagrams in LaTeX? - TeX - LaTeX Stack Exchange, it's mentioned that there's a LaTeX package relying on external dependencies that automatically choose from a few external dependencies depending on availabilities of the system.

This way, pandoc can still provide a native support without committing to a large dependency.

[jgm]

Did some tests. Had good results with rsvg-convert (installed with librsvg), converting to both png and pdf. Bad results with svg2pdf. Bad results with rasterific. Tempted to shell out to rsvg-convert.
I'd rather not use the svg package since that requires -shell-escape.

[Twinside]

Did you had specific bad example files?

[jgm]

This one: https://upload.wikimedia.org/wikipedia/commons/3/30/Vector-based_example.svg +++ Vincent [Aug 10 17 00:50 ]:

…

Did you had specific bad example files? — You are receiving this because you modified the open/close state. Reply to this email directly, [1]view it on GitHub, or [2]mute the thread. References 1. #1793 (comment) 2. https://github.com/notifications/unsubscribe-auth/AAAL5Hh4X4ezeukXRYZpRkBhZ-2Dm0Fvks5sWrZUgaJpZM4DFm9V

[Twinside]

Thanks, I've fixed rasterific-svg to render it correctly, even if it doesn't matter in this case.

[agusmba]

would it make sense to use the same approach for Docx output?

[perlun]

I don't think I'd want to change the default. But it's good to emphasize that people who use SVGs can already use -t context or -t html with -o output.pdf and use these engines, which support SVG already.

Worth mentioning: -t html gives me errors like this, since wkhtmltopdf disables local file access by default since 0.12.6:

  Defaulting to 'my_file_name' as the title.
  To specify a title, use 'title' in metadata or --metadata title="...".
Loading page (1/2)
Warning: Blocked access to file /home/per/some-file.svg
Error: Failed to load about:blank, with network status code 301 and http status code 0 - Protocol "about" is unknown
Printing pages (2/2)                                               
Done                                                           
Exit with code 1 due to network error: ProtocolUnknownError
Error producing PDF.

-t context worked better for me (with librsvg2-bin and inkscape installed) for converting Markdown with embedded .svg images.

[hoijui]

I updated pandoc to version 2.14(.1), and got this same error again.
Reverting back to 2.13 solved it for me.

Can this be reopened, or should I make a new issue?

[jgm]

@hoijui There was a regression in 2.14 with SVGs in creating PDFs.
You can find out about it by searching closed issues. It is fixed in master.