hashtable API

[nyuichi]

No description provided.

[stibear]

What do you think about this?

[nyuichi]

(picrin map)

; map class interface
map?
map-ref
map-set!
map-has?
map-delete!
map-for-each
map-map
map-size
map-empty?
map-clear!

; impls of map
make-hash-map
hash-map
make-tree-map
tree-map

[zeptometer]

When those functions are given, I'm not sure what the word "map" means (map-map will make me mad!). Other words such as "dict" or "table" or etc is more suitable, aren't they?

[KeenS]

How about having
map-size
map-empty?
map-clear

[nyuichi]

@zeptometer

I have some alternatives:

• assoc. I think this is the best alternative among, but it is an abbreviation.
• table. isn't it confusing with other tables (like EXCEL's table)?
• dictionary. Since we have dictionaries as symbol-to-object hash table, we will need to rewrite the sources using old dictionaries; nevertheless, dictionary just suggests tables that looks up with string-like keys, not object-to-object tables.

idea?

[nyuichi]

@KeenS

Indeed!

[KeenS]

@wasabiz @zeptometer
other alternatives are

• hash. This is used in Ruby but as you know this is confusing with hash code, at least, not suitable for tree-map.
• kvs. This exactly explain what it is but is also an abbreviation an initialism.

[nyuichi]

@stibear @KeenS @zeptometer

Hi, all.

Thinking this over for a month, I started to feel like changing dictionaries to object-to-object hashtable is not a bad idea. In the previous reply I complained about that idea by giving two reasons, and now I found they are solvable ones.

1. Since we have dictionaries as symbol-to-object hash table, we will need to rewrite the sources using old dictionaries

I can find only a little uses of dictionary c APIs just by searching current sources codes. It will not require a big change.

1. dictionary just suggests tables that looks up with string-like keys, not object-to-object tables.

Python uses the name "dictionary" to such object. This strongly encourages the change.

Please let me know your positions.

Thanks.

[nyuichi]

Implementation here: picrin-scheme/libpicrin#30

[zeptometer]

I think that is natural extension. Well.. I wish there was way to designate equality: just eq?, eqv?, equal? is enough, like make-hash-table of Common Lisp.

[nyuichi]

@zeptometer

Hmm... I, kind of, agreed. But actually when you use dictionaries only with symbols you can get performance as nice as eq?-based hash tables. I'm not sure this is the case in future version some more eager optimization process say JIT is landed, though.

[KeenS]

@wasabiz I'm for to make dictionary obj-obj map. If you have plan to introduce keywords to picrin, discuss performance then.

[nyuichi]

changed.