Today work has focused on two things; The ctest extension matters, which followed up yesterday’s efforts on the indexing logic in FfiItems’s Extend implementation, and on continuing to work through the Lean book.
The first task that was tackled today was that of finishing up the Extend trait implementation for FfiItems. This I had finished another revision of yesterday but, of course, fresh eyes revealed another flaw in my approach.
The issue was related with correctly interpreting the indices that should have been used in the absence of some kind of item. That lead to a first refactor back into using partition_point, which also had a similar flaw.
I ended up refactoring the whole thing into a far simpler approach that attempts not to use some other kind of item’s index to determine the starting or ending index of a different kind of item. This finally did the trick.
All in all, this took a fairly long while, but I still got a chance to fix all of the spurious errors that had popped up in the codebase after the path aggregate refactor.
One last thing I got to do was to fix the fact I was assuming the Extend trait was the thing I needed. As it turns out, Iterator::collect actually uses FromIterator. Fortunately, no changes were needed.
In terms of Lean work, this time I’ve finished the next section on the book, which had me reading some more about type classes. Most of the stuff I had already read about previously in the language reference.
The one unfortunate downside to this was that some problems in the book are showcased as examples in the language reference. This meant I had to work a tad bit more to both give them another spin and actually solve them.
One such problem I can highlight because I developed it into a set of three activities; Researching the algorithm for instance synthesis, and deriving a formula to hit that limit with a specific data type, followed by experiments.
The research was fairly straightforward to get done because I just had to read through the relevant sections in the language reference and take notes about my interpretation on them.
The formula was also simple to derive, and I used it experimentally to more easily determine the recursive instance search limit without looking at the default value of the option.
None.
I expect the ctest extension matters to continue tomorrow. I have actually finished the next goal I had, which consisted of fixing the spurious errors tied with the path aggregate refactor.
I should now resume the path aggregate synchronization work I had started last week. Lean work will continue as per the usual, which means I will move on to the next section of the book.