|
Just in case it will help, here is a sketch of how I handle it in C++.
Classes and class templates have type Class, one of whose members is an optional list of template parameters (e.g. <T>).
Class template instances have type ClassInst, derived from Class. Rather than generate mangled names, I allow punctuation that can appear in a template argument list to be valid in an internal identifier. That identifier, with each of its constituent names fully qualified, gets added to the symbol table.
|
|
|
|
|
The only trouble with that for me right now is I need to get the CodeDOM to be able to handle that. I may have to resort to namemangling due to the nature of CodeTypeReference but maybe not. It's making my brain hurt right now so I should probably back away. I'll chew on your tactic though.
I've still got to figure out where to keep all those symbols.
I've also been considering importing *all* types from a given namespace import rather than playing go fish, and if i do that, then I'll end up with a table for my generic instances as well. So we'll see.
There's so much here that's still TBD i think i need to outline or something.
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
I hope this is not a programming question disguised as an interjection. Horrors!
The difficult we do right away...
...the impossible takes slightly longer.
|
|
|
|
|
I don't even know how to phrase this level of suffering as a question 
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
Having one object take and action based on the type of another object is usually an anit-pattern. It typically ends up with a bunch of if else statements or even a switch/case statement in some languages: maintenance nightmare for the future.
The answer you seek might be figured out by asking the object itself for the answer, or just telling it to do something.
var answer = this.TellMeWhatINeedToKnow(...);
or
this.DoWhatIsRightForYourType(...); // I Don't Care How You Accomplish It
You might have to support an interface for this to work, or use some other technique like reflection to see if the method is implemented. Does C# support friend methods like C++?
|
|
|
|
|
I don't have control over the objects I'm manipulating and reflection is off the table.
I've reduced some of the switch casing by factoring it into a visitor over the codedom tree but such is the nature of the beast.
The whole point of this project is that the CodeDOM sucks to use, so I'm building a mini programming language on top of it that spits out the constructs for you.
So you don't need to code nasty bizness like this (even wrapped it's bad magic)
var varDecl = new CodeVariableDeclarationStatement(typeof(int),"i",new CodePrimitiveExpression(0));
just to render this:
int i = 0;
Even wrapped you end up with things like:
var varDecl = CD.Var(typeof(int),"i",CD.Literal(0));
Better, but still hard to read.
Slang is a subset of C#
it allows you to declare these constructs in C# so instead of the above you can do like:
var varDecl = SlangParser.ParseStatement("int i = 0;");
to create the same thing.
or even
var varDecl = SlangParser.ParseStatement("var i = 0;");
it will fill in the type because CodeDOM doesn't support var
and of course it can parse more than statements. It can parse groups of C# source files, statements, expressions, namespaces, types and members at the top level as well. Since it uses recursive descent instead of LL(k) it doesn't get confused by that.
Main issue right now is it's not complete and error handling in the parser stinks.
Edited: vardecls are statements, not expressions It's early in the day yet for me. not fully present yet.
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
This question was asked at work yesterday, and a couple of our developers are firmly in the camp that if we were to rewrite the back-end (currently C# with WCF [gross] to handle the endpoint calls and with SQL Server, no ORM, just using Dapper) they would write it in Node with TypeScript.
Conversely, I'm in the camp that if I'm writing anything on the back-end, it's going to be C# / .NET, whether it's a DLL or ASP.NET Core or a simple console app.
So what's you're take? Why would you use one vs. the other for back-end development?
Assuming your 7 year old back-end is written in .NET, if you had to rewrite it, would you switch from .NET to Node?
|
|
|
|
|
I love TypeScript. As a language, it has so much going for it but, and this is a massive but, if I were rewriting a .NET app now, I would go with .NET Core all the way. The new hosting model is faster than Node's; it runs on just about every platform you could possibly want and the new features such as Span<t> means that it can work with much lower code overhead.
|
|
|
|
|
TypeScript isn't a language as much as it is a wrapper around javascript. It's still javascript.
".45 ACP - because shooting twice is just silly" - JSOP, 2010
-----
You can never have too much ammo - unless you're swimming, or on fire. - JSOP, 2010
-----
When you pry the gun from my cold dead hands, be careful - the barrel will be very hot. - JSOP, 2013
|
|
|
|
|
Wouldn't that be true of any language? Say, C or C++ being a wrapper around assembly language, C# and Java being a wrapper around their IL, which in turn is a wrapper around assembly?
|
|
|
|
|
No, it's not the same at all. Typescript simply adds features to an existing language, and is not a language itself. Your other examples are languages unto themselves.
".45 ACP - because shooting twice is just silly" - JSOP, 2010
-----
You can never have too much ammo - unless you're swimming, or on fire. - JSOP, 2010
-----
When you pry the gun from my cold dead hands, be careful - the barrel will be very hot. - JSOP, 2013
|
|
|
|
|
Can you debug typescript as typescript yet, or are you stuck trying to figure out how to map the fugly javascript that it was transpiled to back to your original code?
Did you ever see history portrayed as an old man with a wise brow and pulseless heart, weighing all things in the balance of reason?
Is not rather the genius of history like an eternal, imploring maiden, full of fire, with a burning heart and flaming soul, humanly warm and humanly beautiful?
--Zachris Topelius
Training a telescope on one’s own belly button will only reveal lint. You like that? You go right on staring at it. I prefer looking at galaxies.
-- Sarah Hoyt
|
|
|
|
|
You've been able to debug TypeScript as TypeScript for a long time now.
|
|
|
|
|
cool, does it work in all browsers or need helper plugins?
Not being able to do so was the biggest downfall of the project using coffeescript I was on some years back; and when I asked about typescript a few times before I never got an answer.
Did you ever see history portrayed as an old man with a wise brow and pulseless heart, weighing all things in the balance of reason?
Is not rather the genius of history like an eternal, imploring maiden, full of fire, with a burning heart and flaming soul, humanly warm and humanly beautiful?
--Zachris Topelius
Training a telescope on one’s own belly button will only reveal lint. You like that? You go right on staring at it. I prefer looking at galaxies.
-- Sarah Hoyt
|
|
|
|
|
Dan Neely wrote: cool, does it work in all browsers or need helper plugins?
No plugins required. I'm able to debug TypeScript in VS without problems. VSC - now there I had problems with setting up breakpoints.
|
|
|
|
|
You absolutely can.
Idaho Edokpayi
|
|
|
|
|
Pete O'Hanlon wrote: and the new features such as Span<t> means that it can work with much lower code overhead.
Care to elaborate on that a bit? As in, I can understand memory overhead, so I'm curious how it can affect code overhead?
modified 6-Dec-19 13:50pm.
|
|
|
|
|
I was loose with my terminology here. It should, of course, be memory overhead.
|
|
|
|
|
"We're going to stop speaking English and start speaking French."
If there is a compelling reason to do something then do it. If you're doing something simply because you fancy a change or you think other people are doing it so there has to be a reason, then you shouldn't be in charge of making technical decisions.
|
|
|
|
|
Hey, we finally agree on something. 
|
|
|
|
|
Even with TypeScript, you're still dealing with the Highly Cursed Number type for arithmetic. There is neither checked arithmetic nor unchecked, only cursed, and it is up to you to force it to do the right thing.
|
|
|
|
|
I'd stick with C# . And don't rewrite at all without good reason.
Sadly, where I am now, we're beginning to rewrite all the SSIS (ptui) we've been working on since 2011 in Ab Initio, for no good reason, except for maybe because it costs more and therefore must be better. 
|
|
|
|
|
What's their reason for switching to Node?
Node is JavaScript, which is awful, and TypeScript just makes it look more like C# 
It kind of depends on your environment though.
Personally, I'm using Microsoft all the way down, so switching out anything would make it harder to integrate with everything else.
For example, if you're using Jenkins instead of Azure DevOps, C# integration is a lot harder to begin with and switching to JavaScript may make certain tasks easier.
One reason I'd use Node over .NET Core is to re-use code between a back-end and a front-end (website).
This is actually a scenario that came up recently for a web app we wanted to do with instant front-end calculations that we also had to do back-end for an initial generation for the data.
Node would probably also be a better fit if your data is unstructured (with a MongoDB database, for example).
Using Node could be a hiring decision if you can get plenty of JavaScript developers (because everyone knows JS), but no C# developers.
Other than that I have no reason to use Node instead of .NET Core, they're both multi-platform and lightning fast.
It even seems .NET Core is sometimes even faster than Node!
Node invites the horror that is npm with 10.000+ files in 40 MB for a single package into your back-end 
However, why not use microservices and use both where they best fit? 
|
|
|
|
|
|
Our current ASP.Net WebForms code base requires asmx files so we can communicate with our DAL from javascript.
Our new MVC5 code base is all C#, and we're moving as far away from using javascript (in any of its evil forms) as we can. We also implemented a new DAL that is more generic than the existing one.
We're not using any ORM beyond the EF stuff that comes with your typical MVC project template. To be honest, we'd REALLY like to get rid of EF altogether, but we don't really have time to work on that aspect of the code.
It's hard to decide which I hate more - javascript, or EF.
".45 ACP - because shooting twice is just silly" - JSOP, 2010
-----
You can never have too much ammo - unless you're swimming, or on fire. - JSOP, 2010
-----
When you pry the gun from my cold dead hands, be careful - the barrel will be very hot. - JSOP, 2013
modified 7-Dec-19 6:03am.
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
