I think it's a great idea and it may even be a missing piece in the 'Grow, not Break' approach https://youtu.be/oyLBGkS5ICk?t=1946 , namely to the problem that good names are hard to come by.
Suppose a library author wants to make a breaking change to some function. They change the doc-name of the 'old' function to e.g. foo-old, and reuse foo for the new 'version' of the function. Existing code using this library would work as before, but in the sources the visible name would automagically update to the `foo-old`, thus informing developers that there's a new/better flavor of foo available. This is IMO a far better approach then inventing new names like new-foo, foo2, or my.ns2/foo, for functions that are meant to succeed their legacy variants. Jozef On Thursday, December 7, 2017 at 6:37:10 AM UTC+1, Didier wrote: > > Warning: This is the spawn of an idea, not very well refined, and that is > little extravagant. > > I've been doing some hammock time, and I've been thinking that names in a > programming language are kind of a complecting of two things, the human > readable form, and the machine identifier. What if a function always had > the same name, which never changed, from the moment the function was > created. This would be fine, until a human finds it didn't like the name > anymore, and thus a refactor of the name would occur, requiring to change > all references to the function. This is also true say of a spec, if you > want to change the name of the spec, its a big refactor to update all usage > of it. So var names and spec names are troubling in that if humans want to > refer to it differently, they also break all machine reference to them. > > So I thought, how awesome would it be if each named things in a > programming language would be given a unique machine name, which can be > used everywhere, and the name you saw was just meta-data for the programmer > to have a more human readable/descriptive name. > > The problem is I'm not sure how to do this with a text based programming > language. I can imagine a language where constructs would be first class, > not based on text, and thus all construct could be assigned a unique id and > a name, and the IDEs could easily hide the unique ids and project a view of > the construct with the human name instead. Code would be stored in a > structured format, and obviously a standard editor would not work, and an > IDE of some form would be required. > > So right now, my idea is that maybe you can compromise. If you added > support for Vars and specs, so that they can have like a doc-name. It would > be like a doc-string a bit, but it expects a name instead. That name could > be the human readable name, you could change it freely. The normal var name > or spec name would continue to be used as normal to refer to this var/spec > from other code. At its most basic it means you can have the normal name be > anything, maybe a name you don't like that much, or you could go further > and make it a guid if you want. Then you could make the doc-name the name > you want to use when talking to other people, or when people read the code. > Now IDEs could support this doc-name, so they could show the doc-name in > place everywhere you have code referring to the normal name. They could > auto-complete from doc-name to normal name, etc. > > So an IDE could still kind of hide this for you, and make it appear like > everything is just doc-name pointing to each other, and refactoring would > not require changing all code that refers to it, but actually its just a > change of the doc-name on the var or spec being pointed to, but the IDE > could quickly replace the new doc-name in place of the normal name > everywhere else. > > Where it would be maybe a bit more confusing, is when using an editor that > would not support doc-names to that extent. In those cases, you can ignore > doc-name, consider it just like one more piece of documentation. > > Doc-name could be optional too, so if you plan on working by yourself, or > just in a simple editor, you can ignore this whole thing and work as normal. > > Now maybe this whole thing is solved by having a powerful renaming > refactoring tool that can hunt for all usage and update the name everywhere > in a sound way, but that's harder to do in Clojure, and still breaks when > its a library for example, as you can't just refactor all consumers without > having both access to their code base, and even if you do, its tedious to > pull down everything in your desktop and set it up for such a refactor. > > I'd be interested in thoughts about this, even if its just, that sounds > like a terrible idea :p. > > Thanks. > -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
