Ralf Stephan

That commit was reverted in the same release cycle.

The Maxima part is proposed to be bypassed with https://trac.sagemath.org/ticket/23695

This has worked out and is reasonably fast, although the first command in a Sage session still takes some time because of the imports. Caching not necessary. Still, with creation...

However, in 1e63799 we reverted a use of `dynallocate` because of spurious errors.

Actually there is a performance issue only if the int is used for array indexing.

This attempt revealed repercussions with units: ``` diff --git a/ginac/add.cpp b/ginac/add.cpp index 43c9d89..abc4d1c 100644 --- a/ginac/add.cpp +++ b/ginac/add.cpp @@ -134,8 +134,15 @@ void add::print_add(const print_context & c, unsigned level, bool...

This is not feasible without changing what is expected from `symbolic.Expression._convert()`. However, changing this may be necessary for #71 as well.

I was confused because the Schema syntax didn't have an `id` element but apparently I can just set it nevertheless. Sorry for the noise.

The comments in the original are lost too.

To produce useful failure reports it might be better to have multiple queries, also to optimize query performance and query server load. In general the result of the converter could...