views:

92

answers:

4

I have a little problem with a Pro*C query I'm trying to optimise.

To explain, our application searches for rows in a huge database. These rows exist in several languages and the old code selected a row for each language in an array. Now as these queries are the most time consuming part of our app, I wanted to make only one query which writes directly in an array.

The language codes are 2 letter ISO-639 codes (en for english, fr for french).

Old way (this is only a simplified code to show the intention)

struct ROW arr[MAX_LAN];
struct ROW_IND arr_ind[MAX_LAN];
uint_t LanIdx;
for(LanIdx=0; LanIdx<MAX_LAN; LanIdx++) {
  EXEC SQL SELECT *  /* Don't look at the *, it's for obfuscation only */
      INTO :arr[LanIdx]:arr_ind[LanIdx]
      FROM table WHERE id=:uniqid AND language=:LanCode[LanIdx];
}

I would like to do something like this:

EXEC SQL SELECT *  /* Don't look at the *, it's for obfuscation only */
    INTO :arr:arr_ind
    FROM table WHERE id=:uniqid AND language IN (:LanCodes);

but do not know how I should define LanCodes.

It works with a constant (compile time) list like this

EXEC SQL SELECT *  /* Don't look at the *, it's for obfuscation only */
    INTO :arr:arr_ind
    FROM table WHERE id=:uniqid AND language IN ('en','fr','de');

but this is not useful, as the languages may vary from case to case.

If I write something like

char LanCodes[MAX_LANS*5];
sprintf(LanCodes, "%s", LanCode[LanIdx]);

EXEC SQL SELECT *  /* Don't look at the *, it's for obfuscation only */
    INTO :arr:arr_ind
    FROM table WHERE id=:uniqid AND language IN (:LanCodes);

it works only if there is 1 language code in the string.

So my question is, does anybody know how to make this work? The Oracle documentation is so big, I don't know where to look at. I tried different ways, but none worked.

EDIT Ok, boyz and girlz, I found a solution that works. It's not elegant, it's not advanced but it works well. I put a list of OR clauses in my query and it returns what I need in the form that I need.

EXEC SQL SELECT *  /* Don't look at the *, it's for obfuscation only */
    INTO :arr:arr_ind
    FROM table WHERE id=:uniqid AND (
                language=:v1[ 0] OR
                language=:v1[ 1] OR
                language=:v1[ 2] OR
                language=:v1[ 3] OR
                language=:v1[ 4] OR
                language=:v1[ 5] OR
                language=:v1[ 6] OR
                language=:v1[ 7] OR
                language=:v1[ 8] OR
                language=:v1[ 9] OR
                language=:v1[10] OR
                language=:v1[11] OR
                language=:v1[12] OR
                language=:v1[13] OR
                language=:v1[14] OR
                language=:v1[15] OR
                language=:v1[16] OR
                language=:v1[17] OR
                language=:v1[18] OR
                language=:v1[19] OR
                language=:v1[20] OR
                language=:v1[21] OR
                language=:v1[22] OR
                language=:v1[23] OR
                language=:v1[24] OR
                language=:v1[25] OR
                language=:v1[26] OR
                language=:v1[27] OR
                language=:v1[28] OR
                language=:v1[29] OR
                language=:v1[30]);

It's faster when there is more than 2 languages, so I call this variant or the old one depending on number of languages to fetch.

+1  A: 

You can't do this without Oracle Dynamic SQL. You will have to build your IN clause at runtime and EXECUTE IMMEDIATE. At least you can use Method 1, based on your queries.

DCookie
I'm not sure, with method 3 it would boil down to the same problem. If I make a OPEN C USING :LanCodes will it work? One thing I forgot to mention, is that I cannot build a dynamic statement in a string with changing language codes as this would explode the statement cache of the database.
tristopia
Then I think you need to build a temp table and populate it with your IN values, and join to that.
DCookie
Thank you, but that's even worse (it would take months to get something like that through to our DBAs, we're here in an administration, this would involve a lot of paperwork).I think I will have to live with our current solution (which is not bad btw.). It was only to gain a little on the selects which are the dominant part of the application.
tristopia
Plus, you lose something by having to insert the values in the temp table as well.
DCookie
+1  A: 

Probably this AskTom article can help you.

egorius
Thank you, that looks like something to try out. Follow up on friday to see if it works (and if it is faster).
tristopia
A: 

I have used a table before composed of an ID and a set of rows where the rows are the permutation of the possible values in the "in" list. Then I join to the table based on the ID and it gives me the results I need.

create table permute (
  id number,
  lang char(2)
);
create index permute_p1 on permute ( lang, id );
insert into permute ( id, lang ) values ( 1, 'en' );
insert into permute ( id, lang ) values ( 2, 'en' );
insert into permute ( id, lang ) values ( 2, 'fr' );
...

All you have to do then is pick the correct "ID" value 2, 3, 4 ... and put that into the join.

Philip Schlump
That won't be possible. It's for the EU, so we have to handle 23 official languages (+2 others in the near future), so we have 23! possibilities, that's 25x10^7, the table would be almost as big (or even bigger) as our main database.
tristopia
A: 

... Main String:= 'Select * FROM table WHERE id=:uniqid AND language IN'; -- can split into two to accomadate :uniqd ... Select Language_code into v_string from x_table; loop Copy & Concat v_string to LanCode_String and with ' ', ; end loop; .. Concat Lancode to Main String. .. Prepare and execute the Main String.

Lovell
Yes, that's basically what DCookie suggested. The problem with that approach is that it has a bad influence on the statement cache of Oracle. Overflowing that cache degrades the performance globally (had the effect some time ago for something else). My question was only to save a little on round-trips.
tristopia