Re: [TCLCORE] CFV: TIP#112 and TIP#150

[Donal K. F. <don...@ma...>]

Kevin Kenny wrote:
> The reason that I want to see TIP #112 passed and implemented is
> precisely that I want to begin refactoring the core ensembles to use
> it.  Does every proposal have to have a complete release roadmap for
> its proposed use before it can be approved?
> 
> Refactoring Tcl's own ensembles to use this mechanism is the
> proper subject for other TIPs once we're agreed on the fundamental
> infrastructure.

This is something that I am in complete agreement with.  Such a refactoring,
though a matter that I hope to pursue in the future, is outside the scope of
TIP#112 and *deliberately* so.  It requires as a prerequisite a satisfactory way
of bytecode compiling such things, and possibly a performance study as well.  In
neither case is a TIP required (the compiler API is internal, and performance
studies are not specifications by any stretch of the imagination.)  If/when I
can figure out that making much of the core into an ensemble will improve
things, I'll put forward a separate proposal on the matter which will also
outline a protocol for non-core code to extend those ensembles (or perhaps a
mechanism for preventing it; there are details here that I do not see all the
way through yet.)

OTOH, experiments with the sample implementation (by Joe English) indicate that
what exists is already valuable for user code purely by virtue of it making it
easier for that code to present compound-command-type APIs (and faster too.)  As
such, even if the current TIP presented the entirety of the ensemble API that
got implemented in Tcl 8.5, it would still be a clear step forward.

Donal.
-- 
Donal K. Fellows   http://www.cs.man.ac.uk/~fellowsd/   don...@ma...
-- I'm curious; where does this statistic come from? Does its home, perchance,
   ever see sunlight?     -- Jason A Williams <jas...@co...>