[02:17:34] Al do you think we can speed up Z13397 by splitting the behaviour for types that do/don't serialise? [02:19:29] I'm also a bit concerned at the delay that error handling adds because the test happens on every recursion. I expect we could only do it once then recurse into a helper function that assumes safe bounds. (re @u99of9: Al do you think we can speed up Z13397 by splitting the behaviour for types that do/don't serialise?) [06:49:02] A possibly connected issue. Enums don't yet have read functions, so passing a QID via #function doesn't work. For example, see the third line of [[User:99of9/Sandbox]] where I try to call for Z27962 the chemical symbol of Q1115. (re @David: Thanks for clarifying, Toby. I've created a ticket for this: T405817. (And apologies for the slow response.)) [07:46:01] It’s possible. I think it generally works out that an untyped list is the better form of list argument, but there’s quite an overhead from testing the list’s type, let alone “untyping” a list that doesn’t need it. Now we can catch the error, maybe we can dispatch by the error type and “untype” an unviable list before recursion? (Sorry, I haven’t investigated the message> [07:46:02] new error handling yet.) (re @u99of9: Al do you think we can speed up Z13397 by splitting the behaviour for types that do/don't serialise?) [11:36:22] Yes, that sounds like a good plan, given that the repeated tests should always pass. (re @u99of9: I'm also a bit concerned at the delay that error handling adds because the test happens on every recursion. I expect we could on...) [12:06:01] I can't get Z25866 to complete its tests anymore. [12:14:14] Looks like Z27681 completes in Australian English. I’m guessing the cache got populated with timeouts that were maybe from cumulative rate limits. 🤷‍♂️ (re @u99of9: I can't get Z25866 to complete its tests in time.) [22:08:15] Nudging every failing test worked, but I hope that doesn't have to be done again if the function ever changes! (re @Al: Looks like Z27681 completes in Australian English. I’m guessing the cache got populated with timeouts that were maybe from cumul...)