I’m trying to write an algorithm (the fast marching method for solving the Eikonal equation) that uses a binary minheap with comparisons defined on the nodes of a network. Because I need both the node indices and values all the time, I thought an easy way to implement the algorithm would be to defin…

There’s no benefit to defining your isless method inside the function body, so why not just define it outside? import Base: isless function isless(a::Node, b::Node) ... end function do_stuff() ... end

Oh, I see, your isless is using the T from the local scope. In that case, overloading Base.isless is probably the wrong thing to do, as later calls to isless would keep using the same T in a pretty confusing way. Instead, you can just define an anonymous function and pass that function into binary_…

Thanks, that simplifies things considerably…

I mean…I guess? But think about what’s being implemented here: we have a comparison function that relies on inherently local information (the value of T). Having this isless method automatically attach to the global Base.isless would mean that all comparisons of ::Nodes forever would use that same l…

[image] rdeits: Having this isless method automatically attach to the global Base.isless would mean that all comparisons of ::Nodes forever would use that same local information regardless of how or where that comparison occurs. Couldn’t agree more. But this is exactly the problem ADL solves…

This is the #user:first-steps category. Let’s try not to get sidetracked with developer jargon and theoretical redesigns. @rdeits has it right — you can explicitly pass your comparator to sort. This needn’t be an anonymous function; you can define your own mycompare function and just pass it dire…

100% there is one and only one issue here: this API needs the ability to pass a comparison function. Anything else gets into dynamic scoping, which has historically ended badly and for good reason: it destroys the ability to reason locally about what things mean (or, if your allergic to “meaning” th…

binary minheap using locally defined ``isless''

New to Julia

jlperla May 2, 2018, 4:06am 5

Works here, but I am pretty sure this is another example or where ADL would make this a non-issue (i.e. Function name conflict: ADL / function merging?)

Topic		Replies	Views
New dispatch for isless() does not work General Usage question , numerics , function , multidispatch	2	653	January 4, 2022
Trouble with isless for an immutable type General Usage	4	630	December 16, 2016
Find minimum argument with custom comparator General Usage question	6	1383	October 1, 2020
Meaning of `isless`, `<`, `cmp`, `isequal`, `==`, total order, partial order, unordered Internals & Design	24	2468	January 7, 2018
Structs with custom ordering in a heap New to Julia performance , sort , data_structures	2	1454	September 14, 2019

binary minheap using locally defined ``isless''

Related topics