On Sat, 19 Feb 2022, Hans Hagen via ntg-context wrote:
Hi,
This is related to a bug report for the filter module: https://github.com/adityam/filter/issues/60
Normally, when the same figure is included twice, context only includes the
On 2/19/2022 7:30 AM, Aditya Mahajan via ntg-context wrote: figure once and refers to the included figure the second time. The option `object=no` to externalfigure is supposed to prevent that, but it doesn't work. For example, consider the following example the bug-report uses graphviz as the external program, but I have changed that to context to keep things self contained)
\usemodule[filter]
\defineexternalfilter [graphviz] [ filter=context --purge, output=\externalfilterbasefile.pdf, readcommand=\ReadPDF, ]
\def\ReadPDF#1% {\externalfigure[#1][object=no]}
\starttext
Circle: \startgraphviz \startMPpage draw fullcircle scaled 1cm; \stopMPpage \stopgraphviz
Square: \startgraphviz \startMPpage draw fullsquare scaled 1cm; \stopMPpage \stopgraphviz
\stoptext
Behind the scenes, the first \startgraphviz ... \stopgraphviz environment
writes the content to an external file (\jobname-temp-graphviz.tmp), runs `context --purge \jobname-temp-graphviz.tmp`, which creates `\jobname-temp-graphviz.pdf`. The `\ReadPDF` macro then includees the file (with object=no).
At the next \startgraphviz .. \stopgraphviz, the same process is repeated.
The `\jobname-temp-graphviz.pdf` file is overwritten and the new file should have been included. But it is not. The output contains the first image twice.
I thought that object=no should have prevented the reuse. In fact, if I
compile the file with luatex, I get an error message:
! error: (pdf inclusion): file has changed
mtx-context | fatal error: return code: 256
Any idea why object=no is not working?
'12-two-outputs-temp-graphviz.pdf' the pdf file is kept open, for various reasons:
- performance - the might be different pages to fetch - analysis happens when the image is placed, inclusion happens later
so you try to reopen an already open file (which has not been completely processed)
I see. So, for the filter module this means that one always has to use `cache=yes` (so that each PDF file has a different name). I'll add that to the documentation. Thanks, Aditya