[`memoize-ext`] Wrong hook reference to `begindocument/end`

[jasperhabicht]

prooftrees/memoize-ext/memoize-ext.dtx

Line 851 in fb65220

\hook_gput_code:nnn {begindocument/after} {.}

This should really hook into begindocument/end and probably the not-so internal \@kernel@after@shipout@background should be used instead of \g__kernel_pdfmanagement_thispage_shipout_code_tl.

[cfr42]

Thanks. Commit 0d9515e fixed the hook, which I am apparently entirely incapable of naming correctly. And which is not needed as far as I can tell.

\@kernel@after@shipout@background should be used instead of \g__kernel_pdfmanagement_thispage_shipout_code_tl.

Oh, this is nice. I do not even need to test. On LaTeX, it can run unconditionally.

[cfr42]

Trying now this:

\FirstAidNeededT{memoize}{sty}{2024/12/02~v1.4.1~Fast~and~flexible~externalization}{
  \sys_if_engine_pdftex:T {
    \hook_gput_code:nnn {begindocument/end} {.} {
      \hook_gput_code:nnn {cmd/mmz@shipout@extern/before} {.} {
        \UseName{@kernel@after@shipout@background}
      }
    }
  }
}

[jasperhabicht]

\FirstAidNeededT{memoize}{sty}{2024/12/02~v1.4.1~Fast~and~flexible~externalization}{
  \sys_if_engine_pdftex:T {
    \hook_gput_code:nnn {begindocument/end} {.} {
      \hook_gput_code:nnn {cmd/mmz@shipout@extern/before} {.} {
        \UseName{@kernel@after@shipout@background}
      }
    }
  }
}

This works with PDFLaTeX and LuaLaTeX, but it fails with XeLaTeX which complains about

memoize-extract.py: Python error: 'NoneType' object is not subscriptable.
Traceback (most recent call last):
  File "C:\Users\[...]\AppData\Local\Programs\MiKTeX\scripts\memoize\memoize-extract.py", line 505, in <module>
    width_bp = float(mb[2]) - float(mb[0])
                     ~~^^^
TypeError: 'NoneType' object is not subscriptable

 (test.mmz.log
! Text line contains an invalid character.

Where mb = page['/MediaBox']. Which probably points to missing information on the extracted pages.

However, the above fix can't be used with XeLaTeX either as it generates a similiar endless loop as with LuaLaTeX where the last lines in the log are something like:

Package advice (/mmz) Info: Activated command "\__draw_record_origin:" on input
 line 13.
\openout3 = `memos/E2EC93FD550E56DCDE0E87EE1F0F675E.memo'.

\openout3 = `memos/E2EC93FD550E56DCDE0E87EE1F0F675E-9CE05104B383CF8029692F412DB
6F7A8.memo'.

[1]
No file memos/E2EC93FD550E56DCDE0E87EE1F0F675E-9CE05104B383CF8029692F412DB6F7A8
.pdf
No file memos/E2EC93FD550E56DCDE0E87EE1F0F675E-9CE05104B383CF8029692F412DB6F7A8
-1.pdf
No file memos/E2EC93FD550E56DCDE0E87EE1F0F675E-9CE05104B383CF8029692F412DB6F7A8
-2.pdf

continuing with increasing number suffixes of the PDF file path.

[cfr42]

\FirstAidNeededT{memoize}{sty}{2024/12/02~v1.4.1~Fast~and~flexible~externalization}{
  \sys_if_engine_pdftex:T {
    \hook_gput_code:nnn {begindocument/end} {.} {
      \hook_gput_code:nnn {cmd/mmz@shipout@extern/before} {.} {
        \UseName{@kernel@after@shipout@background}
      }
    }
  }
}

This works with PDFLaTeX and LuaLaTeX, but it fails with XeLaTeX which complains about

memoize-extract.py: Python error: 'NoneType' object is not subscriptable.
Traceback (most recent call last):
  File "C:\Users\[...]\AppData\Local\Programs\MiKTeX\scripts\memoize\memoize-extract.py", line 505, in <module>
    width_bp = float(mb[2]) - float(mb[0])
                     ~~^^^
TypeError: 'NoneType' object is not subscriptable

 (test.mmz.log
! Text line contains an invalid character.

Where mb = page['/MediaBox']. Which probably points to missing information on the extracted pages.

This is a different issue and doesn't depend on \DocumentMetadata or memoize-ext at all. It reproduces if I load just memoize with a standard LaTeX setup. So this looks like an incompatibility between the python extraction script and XeTeX.

I've reported at sasozivanovic/memoize#62 because, as far as I can tell, this is not my fault. The best workaround for now is probably to use the perl extractor instead.