Re: [pdf-devel] Hash module API

[jemarch]

    >  `pdf_status_t pdf_hash_mapkeys (hash_table,
    >                                  pdf_hash_mapkeys_fn_t func,
    >                                  void *client_data)'
    > 
    >      Apply FUNC to each element contained in HASH_TABLE. The prototype
    >      of FUNC should be:
    > 
    >        pdf_status_t func (pdf_hash_t hash_table, char *key, void 
*client_data)
    > 
    >      FIXME: Should we define an order???

   If there exists a situation in wich order is relevant then we can define an
   iterator type over the keys or elements instead of the "pdf_hash_mapkeys()"
   function.

I agree. That solution is much better. It is also consistent with the
list module design.

We should determine an order for the iterators. In a similar function
in the Adobe Acrobat SDK they use the following order schema:

   "Keys consisting solely of digits are enumerated first, in numeric
    order (assuming they are not padded with zeros at the front, which
    will confuse matters). Non-numeric keys are then enumerated in
    strcmp order."

I think would be good to follow the same rule in our implementation.

    >    I don't know how the enum "pdf_hash_value_type_e" is going to be
    >    used yet.
    > 
    > We are not going to use that type, since we are working with generic
    > pointers to store data into the hash table. The definition of
    > `pdf_hash_value_type_e' in gnupdf.texi should be deleted.
    > 

   I agree.

   I'll add the rest of the functions you suggested.

Thanks.