Category talk:Tcl: Difference between revisions

From Rosetta Code
Content added Content deleted
(→‎Language features: Reasoning behind features.)
(→‎Language features: Expand list of language features)
Line 30: Line 30:
<code>safety=safe</code>
<code>safety=safe</code>
:The language, especially in a safe interpreter, has no unsafe operations at all.
:The language, especially in a safe interpreter, has no unsafe operations at all.
<code>paradigms=Imperative, Object-oriented, Event-driven, Reflective, Concurrent</code>
:Tcl supports all of these handily enough. With more detail/justification:
:<code>Concurrent</code>
::The Thread extension is long-established.
:<code>Event-driven</code>
::While [[Tk]] has always been event driven, Tcl has been since 7.5 or 7.6 when it gained the event loop from Tk.
:<code>Imperative</code>
::Tcl is definitely an imperative language.
:<code>OO</code>
::Tcl supports this through many extensions, and natively from 8.6.
:<code>Reflective</code>
::Tcl's had introspection for ages; it's vital for the language's self-tests.

We need to check whether these features are enough; if not, we should update the Language template... —[[User:Dkf|Donal Fellows]] 12:26, 1 June 2009 (UTC)
We need to check whether these features are enough; if not, we should update the Language template... —[[User:Dkf|Donal Fellows]] 12:26, 1 June 2009 (UTC)

Revision as of 13:07, 1 June 2009

Tasks Unlikely to get Implemented

This is a short discussion of the tasks that are marked with the omit template. —Dkf 09:14, 21 May 2009 (UTC)

Constrained Genericity

This is a specialization of the Parametric Polymorphism task, so all comments there apply here too.

Evens Sum To Even

Need the concept of proof in the language for this, and this is really an area that Tcl doesn't address. The formal way to fix it is to write a theorem prover in Tcl, but that'd be a ridiculous amount of work.

Memory Allocation

It's not that we can't do this. It's that we don't. One of the good things about Tcl is exactly not having to deal with this sort of thing. It's a whole layer of stuff that we avoid. But if we were to do it, we'd be looking in terms of:
  • Using string repeat or lrepeat
  • Using SWIG or critcl to write a bridge to a C allocator
  • Showing how to do the classic handle-for-structure extension scheme, as documented in JO's book (and updated for the modern age)

Parametric Polymorphism

Tcl doesn't have static typing, making this task ridiculously trivial/non-applicable. Note also that the task itself states that it only applies to languages with static typing.

Language features

Tcl uses a mixture of pass-by-value and pass-by-reference, and is very good at simulating pass-by-name too. The base language semantics are strictly pass-by-value; this was how everything was done up to Tcl 7.6, and when combined with the fact that it was also string based, it gave the language a (deserved) reputation for being slow. In Tcl 8.0 the language implementation was switched to pass-by-reference, with the entities being semantically immutable objects (the actual code is more nuanced than that, of course); that was a major part of why Tcl sped up with that version. The pass-by-name support is through the upvar command, which allows the looking-up of a variable in one scope and aliasing it to another variable in the current scope. —Dkf 10:16, 31 May 2009 (UTC)

I've selected these features:

expression=dynamic

We use dynamic typing in expr.

compat=duck

We say it's a ducklist if it supports the operations of a ducklist.

checking=dynamic

Our type checks are applied at runtime only. That's when they are enforced strictly.

parampass=value

We always pass parameters by value. We simulate pass-by-reference by passing handles/names and pass-by-name with the help of upvar.

safety=safe

The language, especially in a safe interpreter, has no unsafe operations at all.

paradigms=Imperative, Object-oriented, Event-driven, Reflective, Concurrent

Tcl supports all of these handily enough. With more detail/justification:
Concurrent
The Thread extension is long-established.
Event-driven
While Tk has always been event driven, Tcl has been since 7.5 or 7.6 when it gained the event loop from Tk.
Imperative
Tcl is definitely an imperative language.
OO
Tcl supports this through many extensions, and natively from 8.6.
Reflective
Tcl's had introspection for ages; it's vital for the language's self-tests.

We need to check whether these features are enough; if not, we should update the Language template... —Donal Fellows 12:26, 1 June 2009 (UTC)