Default behavior of slow queries in stdout

[havarddj]

Currently, it seems to me like slow queries are printed to stdout as well as slow_queries.log by default.

Question 1: Is this intended behavior, and if so, is there an easy way to turn it off? One cheap solution is to set slowcutoff = 1000 in config.ini, but that also affects what's put in the log file, which is sometimes undesirable.

The reason I'm asking is that I'm usually not interested in hundreds of lines of:

2025-08-22 15:08:46,742 - SELECT "Ciso", "Clabel", "Cnumber", "abc_quality", "absD", "adelic_genus", "adelic_index", "adelic_level", "ainvs", "analytic_rank", "bad_primes", "class_deg", "class_size", "cm", "conductor", "degree", "elladic_images", "faltings_height", "faltings_index", "faltings_ratio", "iso_nlabel", "isogeny_degrees", "jinv", "lmfdb_iso", "lmfdb_label", "lmfdb_number", "manin_constant", "min_quad_twist_ainvs", "min_quad_twist_disc", "modell_images", "modm_images", "nonmax_primes", "nonmax_rad", "num_bad_primes", "num_int_pts", "optimality", "potential_good_reduction", "rank", "regulator", "semistable", "serre_invariants", "sha", "sha_primes", "signD", "stable_faltings_height", "szpiro_ratio", "torsion", "torsion_primes", "torsion_structure" FROM "ec_curvedata" WHERE "lmfdb_label" = '11.a3' LIMIT 1 ran in  2.2378082275390625s 

whenever I run a request, and if I was, I would presumably find it appearing in slow_queries.log.

Question 2: Would it make sense to change the default behavior (to default: off?)

[roed314]

Besides changing slowcutoff, you can also add silent=False as a keyword argument to the search method. This turns off slow query logging both to slow_queries.log and to stdout (I don't think there's any current way to turn one off and not the other besides doing crazy things like redirecting stdout to /dev/null temporarily while you run a query). Currently this is only an option for the search method, but I'm fully supportive of adding an override to lucky, lookup or other places where this might come up (it sounds like there might be a place in knowls where this is relevant, though that surprises me a bit).

In general, the reason for these slow query warnings is to be annoying so that people try to figure out what indexes are missing. You can find the list of indexes on a table by running db.ec_curvedata.list_indexes() for example. Since there's an index on lmfdb_label, I don't know why that particular query would be slow (you can run the analyze command that's also printed to get some information). So I'm hesitant to make these go away in interactive mode, partially because people are more likely to be annoyed by them there; as far as I know, nobody regularly checks the slow queries logs from beta or production.