| HN Mirror

sure, there's an example at https://bitbucket.org/isti/c-orm/src/4ec76f741d756bf70c2b1b8... but i don't have any text explaining it yet, and it's a bit hard to see what is happening.

but basically there's a python script that parses your struct definitions and then auto-generates a library (this isn't great i know, but what else can you do? if you don't want to use that, you can still use it as an SQL library, but you need to write your own callback functions to do the work of setting values in the struct). that library is

    #include "phonebook.corm.h"

in the linked code. then you can do things like

    static int find_name(isti_db *db, const char *text, name** name) {
      corm_name_select *select = NULL;
      *name = NULL;
      STATUS;
      CHECK(cname.select(&select, db));
      CHECK(select->name(select, "like", text)->_go_one(select, name));
      EXIT;
      if (status == ISTI_ERR_NO_RESULT) status = ISTI_OK; // see NULL name
      if (select) status = select->_free(select, status);
      RETURN;

}

(the STATUS, CHECK, EXIT and RETURN are just macros for the usual "return an int as status and goto exit on error" handling). the snippet above populates the name (a struct typedef) pointer with values (id and name value in this case) from the database. the thing called select is a struct with function pointers that generates SQL select functions (so there's the usual objects-in-C pattern of passing the struct in to the function in "select->name(select, ..." for example). and the SQL being generated is "select * from [table] where name like [value]". it's all escaped correctly to avoid SQL injection.

the API for the database backend is isolated out, but the only current implementation is sqlite.

the complete repo is at https://bitbucket.org/isti/c-orm/src (if you download it and run doxygen you should see better docs - but they're incomplete and not online yet)

if people are interested, email me at andrew@acooke.org and i'll get back to you when it's in a usable state (it mainly needs docs, polishing, and perhaps another database backend - probably postgres).

also, of course, there are many limitations compared to ORM systems. there's no way to retrieve related objects, for example (so if a struct has a pointer to another struct, that pointer isn't retrieved - the best you can do is also store a pointer to the primary key and then make a second call based on that). and currently table and column names must exactly match struct and field names.