The meaning of “meaning” A common theme that comes up when discussing type-piracy/method merging, etc. is that there “can only be one meaning” for a given function without namespace qualifications (I am being careful there to say function, not method). I just want to get some clarification from th…

This seems like a textbook example for why type piracy is discouraged.

The example itself is not important , yes it is pathological … but it shows that you cannot do a “reliable” provable cache, that acts the same as it’s non cached counterpart

@mbauman @jameson I am working on a POC for handling dispatch differently. In a way that gurrantees that if a Module and it dependencies don’t change then it’s binary cache is not changed. Plus as long as you don’t depend on type piracy as a feature , no changes are needed for the code. Plus it …

[image] TsurHerman: recompile everything It doesn’t recompile everything, just whatever needs to change. Sure, that could be a big set of things, but the design doesn’t mandate recompiling everything. The granularity of most makefiles is similarly large; changing a comment in a C++ header cou…

[image] ssfrr: This seems like a textbook example for why type piracy is discouraged. That example is type-piracy, for sure, but that is just because he didn’t put his own type into the dispatch (i.e. extend a Base.+ with a method dispatching on M.MyType is the proper way implement a + in jul…

I have a feeling that a scheme like the following could get rid of a lot of compilation overhead, and would be ugly, but simple: When we need to compile something, we try to bring it into “canonical form” as fast as possible. Then we compute some cryptographic hash of the “canonical form”, and stor…

If or when you do that, provide the future with an easy way to include recognition of return types / types returned, which here would be with respect to change in the returning type. So, when where Julia gets to see what is Just-In-Time or to grok what is more timely yet or allows that uncanny knowi…

[image] jeff.bezanson: I think it would be nice to be able to “seal” a compiled module, in the sense that you can use it but not recompile anything in it even if a method it uses gets changed. That makes me think of the way the Manifest.toml info is used in the new package system to freeze a …

[image] jeff.bezanson: Caching native code for packages would be great but it seems to me we could do it without changing the dispatch system. It would be especially nice if cached compilation results (whether just to the level that the current precompilation works, or all the way to native c…

"Meaning", type-piracy, and method merging

Internals & Design

jeff.bezanson May 31, 2018, 10:35pm 34

No way to know, except the type signature? We don’t specifically detect “type piracy”, but we maybe could if we wanted to, and we certainly use the type signature to decide what to recompile.

Topic		Replies	Views
Julep: Taking multiple dispatch,export,import,binary compilation seriously Internals & Design namespaces	121	9296	May 18, 2018
How to detect/avoid type piracy? General Usage	77	2647	June 14, 2024
MultiFunctions: Context dispatch and binary cache Internals & Design proposal , scope	9	2385	July 16, 2018
Binary Caching and Pre-compilation .. again Internals & Design precompilation	7	3661	July 22, 2022
Is dispatch on another type's parameter (::NotMyType{MyType}) piracy? General Usage	32	812	May 29, 2024

"Meaning", type-piracy, and method merging

Related topics