You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In 1.10.7 and 1.11.2, one could autocomplete, for example, using DataFrames; show(df, allrTAB and get show(df, allrows=. This no longer works on 1.10.9 and 1.11.4.
Kwarg completions on functions with more than 40 methods was subsequently disabled due to #54131. But that case is quite extreme! An MWE here that emulates what PyCall is doing is:
struct F; end
Base.getproperty(::F, ::Symbol) = Any[cos, sin][rand(1:2)]
f =F()
f.bar(<TAB>
In short, this asks for tab completion on (::Any)(...) — that is, every method in the system! All 30k+ of them. So, yeah, that's gonna be sluggish.
The fix in #56963 used MAX_METHOD_COMPLETIONS to limit this. That's set to 40, based upon a display limitation. I think it'd be reasonable to increase this to something like 500, which would cover nearly all functions. In the examples above, finding the allrows kwarg for show — which has 304 methods with DF.jl loaded on 1.11.2 — has no noticeable latency.
Could we increase the limit here?
The text was updated successfully, but these errors were encountered:
In 1.10.7 and 1.11.2, one could autocomplete, for example,
using DataFrames; show(df, allr
TAB and getshow(df, allrows=
. This no longer works on 1.10.9 and 1.11.4.Kwarg completions on functions with more than 40 methods was subsequently disabled due to #54131. But that case is quite extreme! An MWE here that emulates what PyCall is doing is:
In short, this asks for tab completion on
(::Any)(...)
— that is, every method in the system! All 30k+ of them. So, yeah, that's gonna be sluggish.The fix in #56963 used MAX_METHOD_COMPLETIONS to limit this. That's set to 40, based upon a display limitation. I think it'd be reasonable to increase this to something like 500, which would cover nearly all functions. In the examples above, finding the
allrows
kwarg forshow
— which has 304 methods with DF.jl loaded on 1.11.2 — has no noticeable latency.Could we increase the limit here?
The text was updated successfully, but these errors were encountered: