LaTeX reader: filename without extension in \includegraphics

[bbc2]

The following foo.tex compiles with pdflatex but not pandoc:

\documentclass{article}
\usepackage{graphicx}
\title{Foo}
\begin{document}
\includegraphics{bar}
\end{document}

Directory layout:

.
├── bar.png
└── foo.tex

Pandoc command:

> pandoc --self-contained foo.tex --output foo.html
File bar not found in resource path

Expected result: no error.

Without --self-contained, the output is also unexpected (missing .png extension):

<p><img src="bar" alt="image" /></p>

Version:

> pandoc --version
pandoc 2.2.3.2
Compiled with pandoc-types 1.17.5.1, texmath 0.11.1, skylighting 0.7.3
Default user data directory: /home/bertrand/.pandoc
Copyright (C) 2006-2018 John MacFarlane
Web:  http://pandoc.org
This is free software; see the source for copying conditions.
There is no warranty, not even for merchantability or fitness
for a particular purpose.

[jgm]

How does latex behave? Does it have a certain list of extensions that it tries when none is specified? In what order?

[bbc2]

I just had a look at the manual available at https://ctan.org/pkg/graphicx but it doesn't say anything about default extensions. However, it mentions pdftex.def which is defined in the graphics-def package. In that file, I found https://github.com/latex3/graphics-def/blob/dc7e8133c0410bb8836e43f99b78274ad7fc24aa/pdftex.def#L522-L525:

\def\Gin@extensions{%
  .pdf,.png,.jpg,.mps,.jpeg,.jbig2,.jb2,%
  .PDF,.PNG,.JPG,.JPEG,.JBIG2,.JB2%
}

If both .png and .PNG are removed from the list, then pdflatex can't find bar.png in my earlier example, which seems to confirm the purpose of that definition. A quick strace shows that those extensions are tried in the same order as listed.

(For some reason, if only .png is removed, then a .png still works. After failing to open bar.PDF and bar.PNG, it tried bar.png anyway.)

The manual also describes options and commands that can modify what extensions are tried:

• The \DeclareGraphicsExtensions command adds extension candidates.
• The \DeclareGraphicsRule command changes how some file extensions are handled, which can involve executing a shell command.
• The type, ext, read and command options of the \includegraphics command have the same meaning as in DeclareGraphicsRule.

I suppose that emulating all those commands would be too much, but just maybe supporting the default extensions and perhaps even \DeclareGraphicsExtensions would be nice?

[eilers]

I think it would be good enough if pandoc provides a command line parameter that sets the fallback extension if it is missing in the filename. I think in most cases that should solve the issue.

[jgm]

@eilers You mean, like --default-image-extension? (See the manual)

[jgm]

It should be possible to modify the LaTeX reader to check existence for files, when the filename has no extension, using this order:
.pdf .png .jpg .mps .jpeg .jbig2 .jb2 (and their capitalized variants)
If a match is found, we could store the filename + the found extension in the Image element.