A week or so out from release 9.0.0

[rocky]

I think we are about a week or so out from of Mathics3 9.0.0. There are still some small bugs to be addressed.

In preparation for that I just pushed a new Docker image. The announcement on Mathics3-omnibus has more information on this image and what's in the next release.

[rocky]

I think we are about a week or so out from of Mathics3 9.0.0. There are still some small bugs to be addressed.

In preparation for that I just pushed a new Docker image. The announcement on Mathics3-omnibus has more information on this image and what's in the next release.

It's been a couple of weeks, so let me give an update on the next release.

@mmatera has been working on the correct selection of a Rule when several match, and correcting the Set builtin functions, Set, SetDelayed, UpSet, UpSetDelayed, TagSet, and TagSetDelayed, so that they handle conditions properly.

This is very useful and important work. The plan for work after the 9.0.0 release is to start revising Boxing and formatting. I imagine that will make heavy use of Rules and different Set functions, so we may as well get the correction out of the way now.

[mmatera]

Going over assignment (*Set*) evaluation, I hit some long standing issues related to patterns and rules

#1441: Conditions in the RHS of an expression must stay at the RHS, but must be considered when rules are ordered for application priority. Also, this means we have to check if in a Definition values, there are rules with equal LHS, but with different conditions in the RHS, to avoid overwrite inequivelent rules.
#1442 This is the main rework of the assignment routines implementation, to make they work closer to WMA
#1447 -> #1449 These two PRs are related to the way in which expressions and rules are sorted. I did it in two stages: first moving the relevant code to a specific module. Then, changing the way the sort keys are stored, to make it a little bit more transparent, following the design pattern used for Attributes.

There are more adjustments to do: just as an example, how to process Verbatim[...] and HoldForm in the LHS of an assignment expression, improve the compatibility with the pattern-like sort for conditional expressions, and how Conditional expressions as handled as return values inside modules, but these should be more or less straghtforward changes once the ones in the listed PRs be merged.
These adjustments should make Mathics more compatible with WL packages, that uses these special assignment forms.

Also, these changes would make it possible to reimplement the MakeBoxes rules as WL rules instead of writing them as Python code, as @rocky suggested some time ago.

[rocky]

I think we are about a week or so out from of Mathics3 9.0.0. There are still some small bugs to be addressed.

In preparation for that I just pushed a new Docker image. The announcement on Mathics3-omnibus has more information on this image and what's in the next release.

It's been quite more than a couple of weeks, more like a month, but at this point, all of the core pieces have been updated and are available from both PyPI and GitHub.

But I think (or hope) it has been worth it. There are some fundamental improvements that move us closer to WMA conformance.

Some of the important changes:

• Significant progress on box operator parsing and scanning escape sequences properly
• Rule selection for pattern matching more closely follows WMA
• Assignment operators isolate left-hand-side references.
• Interrupt handling added
• Debugging capabilities improved
• We start to track locations which is used in errors and in debugging better

What remains to be updated are the conglomerates Mathics3 omnibus and updating pyodide. I will try to update these soon.