Some remarks on the GAP integration:

We (= GAP) have a new helper function CALL_WITH_STREAM which you could use to rewrite your capture_stdout with pure GAP. As an example, in GAP.jl (Julia interface to GAP), I use these two helpers; adapting that code to match capture_stdout is straight-forward. Note that I use SetPrintFormattingStatus there to avoid diffs caused by the width of the user's terminal; but if you want to 100% faithfully reproduce what GAP does, then leave that out.

BindGlobal( "StringDisplayObj",
function(obj)
  local str, out;
  str := "";
  out := OutputTextString(str, false);
  SetPrintFormattingStatus(out, false);
  CALL_WITH_STREAM(out, Display, [obj]);
  CloseStream(out);
  return str;
end );

BindGlobal( "StringViewObj",
function(obj)
  local str, out;
  str := "";
  out := OutputTextString(str, false);
  SetPrintFormattingStatus(out, false);
  CALL_WITH_STREAM(out, ViewObj, [obj]);
  CloseStream(out);
  return str;
end );

This is of course in the end equivalent to what capture_stdout does, but while CALL_WITH_STREAM is also undocumented, it'll be more stable than calling internal kernel functions like OpenOutputStream. We could also eventually offer a documented, high-level version of it as CallWithStream; the main reason that's not yet there is that I first wanted to validate the design. If Sage also finds it useful, that would be a strong motivation to provide a fully documented CallWithStream in GAP 4.13. In addition or alternatively, functions like StringViewObj or StringPrint could be provided, but there it is already less clear that we can agree on a common version (see SetPrintFormattingStatus ...) ?

• Another point: you could replace your uses of GVarName and AssGVar by GAP_AssignGlobalVariable (the performance overhead of re-calling GVarName should be miniscule; if hold_reference ever is called in a tight loop, that code likely has more serious problems than that...

• alternatively just remove hold_reference (nothing uses it?)

• remove NewBag and CallbackForAllBags from gap_includes.pxd ?

• there are a lot of other things that could be replaced by "official" libgap API usage, e.g. IS_INT -> GAP_IsInt, IS_INTOBJ -> GAP_IsSmallInt, SUM -> GAP_SUM, VAL_MACFLOAT -> GAP_ValueMacFloat, etc. -- most of these were added specifically for use by Sage years ago but apparently are not actually used by Sage? Overall I would expect that a majority of the imports from outside libgap-api.h can be replaced. If someone wants to work on that and needs help to figure out what replaces what, I'll be happy to assist

Originally posted by @e4745025-e0be-4c7e-a474-0267c8aa7406 in #34391 (comment)