I was using NLsolve and couldn’t get type stability in my model no matter how. I’m new to Julia, so I thought I wasn’t fully understanding how type stability works. But I think this is part of a bug that dates from at least 2016 (see #15276 ). Just wanted to confirm it, since it basically determines…

First I would make a difference between: a bug: incorrect behavior performance issue: correct behavior but slow performance The case we observe here is performance issue in my opinion. In particular, what we discuss is the performance of Julia in its current version (I tested it on 1.8.0 - in fut…

That for sure an interesting case :slight_smile: I think the perception of this depends a lot on what your previous programming experiences are. Coming from a C++ background, I’m more surprised that result1 works at all, simply since ff1 is used before being defined. But there are many languages wh…

[image] alfaromartino: since it basically determines that the order and place where you put functions is crucial for type stability. Thanks for your response! My point is that my case is different: I’m not calling any function. I should have rephrased this and said “since it basically determ…

Mh, I think it is very similar, since ff1 is essentially a variable name which is bound to a function. The datatype is Function instead of Int64 but we still use a variable name to refer to it and it could be undefined or overwritten. I admit it is a bit strange to think of functions as something wh…

[image] alfaromartino: If it’s not a bug, I’d appreciate whether you could explain to me how I should think about it (in particular how to think about CASE 1 and CASE 2 below, which completely puzzles me!!!). Case 1 is as documented in the performance tips, while the situation with Case 2 is …

[image] nsajko: you just got lucky with Case 2. Lucky? What do you mean? I’m defining exactly the same functions in all cases and not executing any of them. I’m only changing the order in which I define functions before any execution. And then the execution performs the calculations followin…

A bug is when software behaves worse than documented. The situation is the opposite here.

Where exactly is this documented? You even said in the post you deleted that this is a known bug [image] Ominous type instability Performance This seems to be a known issue:

[image] alfaromartino: Where exactly is this documented? Performance tips, you provided the link. You even said in the post you deleted that this is a known bug Well I deleted the post because I changed my mind. I don’t think the issue you discuss has anything to do with the bugs I linked…

[image] alfaromartino: But I think this is part of a bug that dates from at least 2016 (see #15276 ). IMO it is this bug, effectively. I would just avoid that kind of construct (avoid nested functions and/or pass them as parameters, as in case 4). Even if that “worked” having a function defin…

Thanks @lmiq . I can see why nested functions could make the code hard to read. My concern was for two specific cases. I thought that wrapping the main script into a function was sometimes good for performance (see here for example ). Do you think this could create some issues in this respect? My …

Is this a bug? Nested Functions and NLsolve

General Usage

alfaromartino August 21, 2022, 10:01pm 9

Where exactly is this documented? You even said in the post you deleted that this is a known bug

1 Like

Topic		Replies	Views
Can someone explain closures to me Internals & Design performance , type-stability , closure	13	3474	November 2, 2023
Immediately executing closure as function barrier General Usage closure	16	956	September 30, 2022
Unrecommended style of programming in Julia for type stability? General Usage type-stability	17	2314	December 12, 2017
Define data-dependent function, various ways New to Julia closure , functors	34	2524	November 2, 2020
Type instability of nested function New to Julia	5	891	March 12, 2021

Is this a bug? Nested Functions and NLsolve

Related topics