Re: table library API change proposal

["H. Diedrich" <hd2010@...>]

Hi Benoit,

I think the extension is a good proposal.

Benoit Germain schrieb:
> The current table library isn't convenient to manage FIFOs
> efficiently.
But your proposal, incl the FIFO problem, is efficient probably only 
under special circumstances: rare in/out, small queue.

Basically, you are implementing a FIFO architecture on top of something 
that isn't made for it. This will, to my understanding, scale badly.

If the API was smoother for the task, it would still leave you with the 
problem that you are extensively shifting elements around for every 
access, when what you really want is a linked list with a view on the 
current head and tail.

Since you don't care for the table keys in a FIFO but the table can't be 
told to ignore them when doing all the internal shifting, your solution 
may be to use the keys, not shift the tables at all, i.e. not use remove().

Not sure, also not to contradict your principal request. But it may be a 
pointer to why remove is not more powerful currently, because it's 
costly internally.

Henning