Prev: Recommended parser / lexer generator?
Next: Bug in sqlite3?
From: cattaghia on 25 Apr 2010 16:51
hehehe! :) Well, anyway, now I think that "string is list" is not so
reliable.

But why does "dict exists" raise an error? This is an undocumented
behaviour, and the purpose of the command is exactly to avoid an error
from a future "dict get"...

Fabricio
From: tom.rmadilo on 26 Apr 2010 01:28
On Apr 25, 1:30 pm, "Donal K. Fellows"
<donal.k.fell...(a)manchester.ac.uk> wrote:
> On 25 Apr, 21:03, cattaghia <cattag...(a)ig.com.br> wrote:

> > As Gerald said, "most strings are lists"... I can't see how Tcl can
> > state that "Tom is a cat" is a string and not a four-elements list.
>
> Oh, but it *is* a four-element list! And a string as well at the same
> time. Types are merely a matter of perspective. "So what I told you
> was true... from a certain point of view." "A certain point of view?"
> "Luke, you're going to find that many of the truths we cling to depend
> greatly on our own point of view."

Is it even possible to test if a sting could be a dict? I've never
seen such a flexible structure as a tcl dict, and I don't think this
is a bad thing at all. But if we impose structure by feeding a string
to a list command, or an array command, dict seems to require/allow a
recursive dose of the usual structural interpretations.

What that means to me is that you shouldn't be guessing (with a
generic test) if a sting is a dict, because this really tells you
nothing. Any particular dict command could be correctly invalid, or
incorrectly valid given a valid dict, so a test for a valid dict is
mostly useless.
From: Donal K. Fellows on 26 Apr 2010 04:11
On 26 Apr, 06:28, "tom.rmadilo" <tom.rmad...(a)gmail.com> wrote:
> Is it even possible to test if a sting could be a dict? I've never
> seen such a flexible structure as a tcl dict, and I don't think this
> is a bad thing at all. But if we impose structure by feeding a string
> to a list command, or an array command, dict seems to require/allow a
> recursive dose of the usual structural interpretations.

Syntactically, a dictionary looks like list (any list) subject to one
or two extra conditions. The *strict* condition of well-formedness is
that the list must have an even number of elements. The loose
condition is that the key elements (i.e., first, third, fifth, etc.)
should all be unique within a particular dictionary; this condition is
never enforced by anything parsing a dictionary, but instead only the
last such key/value pair is used. (Why? Ultimately because that was
easiest to implement.) My [string is_dict] from earlier in this thread
enforces the strict conditions of looking like a list and having an
even number of elements, on the assumption that it is being used as a
check to see whether the value can go into a [dict] operation without
causing a syntax error. (Canonicality - which for dicts is equivalent
to canonicality for lists plus the no-dup-key condition - is a much
stricter and more expensive requirement, and one that the string from
earlier in this thread would still pass.)

> What that means to me is that you shouldn't be guessing (with a
> generic test) if a sting is a dict, because this really tells you
> nothing. Any particular dict command could be correctly invalid, or
> incorrectly valid given a valid dict, so a test for a valid dict is
> mostly useless.

Yes. It's probably not a good idea to accept a dictionary directly
from general user input.

Donal.
From: cattaghia on 26 Apr 2010 18:31
Hmm. So I presume that "dict exists" checks this pairs "strict
condition" on its argument before actually looking for a certain key.
If the list does not have an even number of items, boOoOom.

Not bad, really. But should not the documentation tell about the
possibility of an error if a bad-formatted dict is thrown on it?


Fabricio
From: tom.rmadilo on 26 Apr 2010 20:23
On Apr 26, 3:31 pm, cattaghia <cattag...(a)ig.com.br> wrote:
> Hmm. So I presume that "dict exists" checks this pairs "strict
> condition" on its argument before actually looking for a certain key.
> If the list does not have an even number of items, boOoOom.
>
> Not bad, really. But should not the documentation tell about the
> possibility of an error if a bad-formatted dict is thrown on it?

I'm not a real user of dict, but I have run a number of experiments.
Basically it is a portable, hierarchical Tcl array. But only the top
level can be tested for an odd number of values !(len & 1). For
instance dict x: {a {b c} d {e f g} h {i j k l}} passes the top level
test. Path {a b} exists, path {d e} is invalid, path {h k} is valid.
If you [dict lappend x d m], then path {d e} is now valid and returns
"f". Not clear to me is why the [dict lappend x d m] doesn't create a
dict {a {b c} d {{e f g} m} h {i j k l}}, but that is the way it
works.


First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: Recommended parser / lexer generator?
Next: Bug in sqlite3?