[05:29:10] Because it is a very low level function, and I didn't want to handle lexemes or items just to indicate the person and number. No human user is intended to directly use the function anyway (re @u99of9: Why is "subject person and number" a natural number?) [05:35:44] Basically they're corner cases like "vorresti esserti pettinata", that instead the function now generates "ti vorresti avere pettinato". The handling of unstressed pronouns is very complicated, since they are full of weird grammatical rules, exceptions, corner cases, exceptions to the exceptions... and when modal verbs are present too, everything becames even [05:35:44] more messed up (re @ [05:35:45] amire80: In what way?) [13:48:39] I made a start… [13:48:39] [[Wikifunctions:Type proposals/HTML fragment structure]] [13:48:41] It basically extends and generalises Z33470 so that it’s conceptually closer to the arguments to this, for example, but in a single object without the function calls. [13:48:42] https [13:48:42] //www.wikifunctions.org/view/en/Z33470?call=%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z33470%22%2C%22Z33470K1%22%3A%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z27854%22%2C%22Z27854K1%22%3A%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z32234%22%2C%22Z32234K1%22%3A%5B%22Z1%22%2C%22West+Bay%22%2C%22%2C+%22%2C%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z24766%22%2C%22Z24766K1%22%3A [13:48:42] [13:48:44] %7B%22Z1K1%22%3A%22Z6091%22%2C%22Z6091K1%22%3A%22Q55231693%22%7D%2C%22Z24766K2%22%3A%22Z1113%22%7D%5D%7D%7D%2C%22Z33470K2%22%3A%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z21394%22%2C%22Z21394K1%22%3A%5B%22Z6%22%2C%22span%22%5D%7D%2C%22Z33470K3%22%3A%7B%22Z1K1%22%3A%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z883%22%2C%22Z883K1%22%3A%22Z6%22%2C%22Z883K2%22%3A%22Z1%22%7D%2C%2 [13:48:44] 2K1%22% [13:48:45] 3A%5B%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z882%22%2C%22Z882K1%22%3A%22Z6%22%2C%22Z882K2%22%3A%22Z1%22%7D%2C%7B%22Z1K1%22%3A%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z882%22%2C%22Z882K1%22%3A%22Z6%22%2C%22Z882K2%22%3A%22Z1%22%7D%2C%22K1%22%3A%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z21394%22%2C%22Z21394K1%22%3A%5B%22Z6%22%2C%22lang%22%5D%7D%2C%22K2%22%3A%5B%22Z6%22 [13:48:45] %2C%22e [13:48:47] n%22%2C%22-%22%2C%22AU%22%5D%7D%5D%7D%7D (re @u99of9: That sounds like a whole new Type. Do you want to write it up?) [15:44:34] I haven't looked at these specific implementations, but I should mention that we are reverting one aspect of "Orchestrator V2". Prior to V2, arguments going to code implementations were generally fully resolved. V2 changed that, rather abruptly, so that they were only resolved at the top level; nested parts of arguments didn't necessarily get resolved. The reversion [15:44:35] means that [15:44:35] once again, arguments going to code implementations will be fully resolved. Sorry, I'm not sure exactly when this reversion will happen, but soon. Will check on this. Also, there will be more communications related to this issue of degree of resolution. (re @Al: It’s the indirection, I think. The problem affects only a few implementations where the argument reference [15:44:35] is not [15:44:36] resolved up fr...) [18:10:54] Thanks for keeping us updated. Hopefully we can eventually find a happy place in the middle, but it sounds like we should expect an initial performance hit, whether or not we notice it. (re @David: I haven't looked at these specific implementations, but I should mention that we are reverting one aspect of "Orchestrator V2". ...) [21:03:14] Thanks, Al . That reversion that I mentioned just above already happened today. By "performance hit" I assume you mean tests that might break because they had already been adapted to account for V2 changes. I'm not aware of any specific performance hits, but yes that is possible. This reversion was motivated primarily by T419789, for which you had disconnected Z14403. [21:03:14] I felt [21:03:15] that was particularly unfortunate because of the importance of Z14310. I just visited Z14403 and all of its tests flipped from failing to passing. Very happy to see that. So feel free to re-connect it, or let me know if you would like me to do so. [21:16:26] Yes, it was rather. Fortunately the v2 snappiness made the recursive compositions fairly competitive. I’ve reconnected Z14403. [21:16:27] The performance hit I mentioned would be from resolving beyond the first level. As I said, we may not notice. Thanks again. (re @David: Thanks, Al . That reversion that I mentioned just above already happened today. By "performance hit" I assume you mean tests t...)