| 28 Jul 2025 |
 Sergei Zimmerman (xokdvium) | Microbenchmarks with google/benchmark? | 14:16:08 |
 John Ericson | well my argument above is microbenchmarks would kinda miss the point | 14:16:42 |
| John Ericson | yes, parsing does become more expensive | 14:16:48 |
| John Ericson | but other things become less expensive | 14:16:53 |
| John Ericson | and it should be some O(drvs) thing either way | 14:17:06 |
| Sergei Zimmerman (xokdvium) | In reply to @Ericson2314:matrix.org
well my argument above is microbenchmarks would kinda miss the point But that change would affect store derivation instantiations, right? Or is that not the case? | 15:54:37 |
| John Ericson | what do you mean by instantiations? | 16:05:59 |
| John Ericson | creating them in the evaluator? | 16:06:05 |
| 29 Jul 2025 |
 Mic92 | John Ericson: Sergei Zimmerman (xokdvium) https://github.com/NixOS/nix/pull/13569 | 07:42:09 |
| Mic92 | Microbenchmarks for derivation parsing | 07:42:58 |
| John Ericson | thanks Mic92! | 14:58:04 |
| 31 Jul 2025 |
|  Sammy (It/Its) left the room. | 10:11:40 |
| 29 Jul 2025 |
| John Ericson | https://github.com/NixOS/nix/issues/13570 issue I am curious about | 14:58:08 |
| Mic92 | Left a comment. | 15:17:42 |
| John Ericson | thanks! | 15:24:40 |
| John Ericson | didn't think of that angle | 15:24:43 |
| John Ericson | Mic92: any idea where we would get a nice meaty real-world structured attrs derivation? | 20:17:00 |
| John Ericson | I'll need that to test my PR | 20:17:26 |
| John Ericson | nevermind | 20:35:30 |
| John Ericson | firefox.overrideAttrs (_: { __structuredAttrs = true; }) that works | 20:35:44 |
| John Ericson | oh nevermind i grepped wrong, it was already structured attrs! | 20:38:06 |
| John Ericson | Mic92 Sergei Zimmerman (xokdvium) OK yeah firefox parses 4 times slower | 20:39:21 |
| Sergei Zimmerman (xokdvium) | In reply to @Ericson2314:matrix.org
Mic92 Sergei Zimmerman (xokdvium) OK yeah firefox parses 4 times slower So doing the lazy thing is worth it certainly. | 20:40:10 |
| John Ericson | OK! | 20:40:17 |
| John Ericson | Sergei Zimmerman (xokdvium): I wonder if I should use mutable since it is morally pure | 20:41:03 |
| John Ericson | not good for concurrency const | 20:41:13 |
| Mic92 | In reply to @Ericson2314:matrix.org
Mic92: any idea where we would get a nice meaty real-world structured attrs derivation? Anything in stdenv has to be parsed/generated multiple times. Bootstrapping in general takes most of the time in any eval flamegraph I have seen so far. | 20:41:19 |
| John Ericson | but is good for "are we changing it?" const | 20:41:21 |
| Mic92 | So this Code makes parsing 4x slower? https://github.com/NixOS/nix/commit/7f0ff4957e634de8aa2edb3cac5e79375c605148 | 20:46:04 |
| Mic92 | Surprising because the throughput of our current parser is slower than what I expect a JSON parser to be able to pull off. | 20:47:48 |
