Dave M Booster Grant Report
Thu, 11-May-2023 by
Saif Ahmed
edit post
< is a useful summary of deparse. Brian DeFoy has aso gives useful [insight](https://www.effectiveperlprogramming.com/2010/05/use-bdeparse-to-see-what-perl-thinks-the-code-is/) on how Deparse can de-obfuscate Perl code, pick up errors etc. This has taken around 30 hours of work. It certainly highlights a feature I knew very little about at all.
### Rework
Dave has done a complete reimplementation of scope internals: by scope we mean blocks of code which relate to a particular context. The context may represent the program as a whole, a subroutine, or a loop. There is reduction of reliance on recursion, the internal code is made more readable, and steps to prevent buffer overrun, improve sprintf, list assiugnment etc.
### Stand up and be counted
While fixing specific issues that have been raised is important, there are key components that may be repsonsible for multiple failures and many issues. These are difficult to spot. One of these relates to the way Perl manages data and the memory locations where the data is kept. [post by David Farrell](https://www.perl.com/article/the-trouble-with-reference-counting/#:~:text=Perl%20uses%20a%20simple%20form,belong%20to%20the%20block%20scope.) highlights how complex this problem is. When variables go out of scope and no longer required they are (or should be) cleaned up, and the memory used freed. If this space are not correctly freed, then there is poetntial for memory leakage and the routines taking up more and more space through each call; conversely innapropriate removal of references mean that a variable that is expected to be accessed can not be. Both lead to failures and are potentially responsible for many “unfixable errors” in Perl modules. But reference counting is not simple. Many objects nest references deeply, and further more, Perl owes some of its flexibility to allowing programming behaviours that may be frowned on in other languages. The price of this flexibility is a rather large amount of stress on Perl core.
So to identify these problems Dave has created a wrapper for over 250 functions. This has led to a feature that allows references to be counted and traced to allow identifications of points of failure. This wrapper can be optionally activated with a minor amount of slowdown, but once the robustness of the code has been verified, can be removed. This work is enormous, and has taken close to 200 hours of work, as such modifications must be both reversible and not break the working of Perl. This extra activity has thus eaten up a lot of the grant very early in the term. A rather big risk which, Dave himself admits, may not be completely successful; at the same time, this one mega-effort has the potential to transform the robustness of Perl core.
### Conclusion
This is a over-simplistic overview of a few small components of the work that our Core developers, including Dave Mitchell, Tony Cook and many others, do to make Perl continue to be relevant. Undoubtedly this is a task that requires a depth of knowledge that is not ubiquitous in the community that depends on Perl. It is great to see that the Grants system of The Perl and Raku Foundation being used so energetically to continue improving Perl, adding new features, making it more robust and fixing things that were considered unfixable before, so enormous debt of gratitude also has to go to our sponsors.
]]>
Comments (0)
