MySQL calling a procedure with select statements from a trigger - “Not allowed to return a result set from a trigger”

Question

Here's one for the SQL wizards. Similar questions have been asked before and I have some idea what the problem is but I'm having trouble rewriting the select statements inside my procedure to avoid returning a result set, and returning the global vars my trigger wants.

For a bit of background, what I'm trying to accomplish here is to use a group of triggers to call procedures which will calculate the percentile of a value relative to other rows in the table with the same gene_id, on insert or update of rows.

(using PMA delimiter is set to |)

The procedure to calculate the percentile value works fine

DROP PROCEDURE IF EXISTS gecg_relative_expr_compendium |

CREATE PROCEDURE gecg_relative_expr_compendium( IN in_gene_id varchar(24), IN in_fpkm double, OUT percentile float(4,2) ) 
BEGIN 
    SET @row := 0;
    SELECT fpkm, ( (rank / @row) *100 ) AS percentile
    FROM (
        SELECT fpkm, @row := @row +1 AS rank, gene_id
        FROM gene_expression_cufflinks_gene
        WHERE gene_id = in_gene_id
        ORDER BY fpkm ASC
    ) AS p
    WHERE fpkm = in_fpkm;
END |

I can call it like so:

CALL gecg_relative_expr_compendium('ENSG00000251948', 19367800, @percentile)

Returns:

fpkm    percentile
19367800    100.0000

And my trigger:

DROP TRIGGER IF EXISTS insert_gecg_relative_expr |

CREATE TRIGGER insert_gecg_relative_expr

 BEFORE INSERT ON `gene_expression_cufflinks_gene` FOR EACH ROW  
 BEGIN  
    CALL gecg_relative_expr_compendium(NEW.gene_id, NEW.fpkm, NEW.fpkm_percentile_compendium) ;
 END  |

When I insert a row triggering the trigger, it fails (irrelevant columns trimmed for clarity)

 INSERT INTO `mctp`.`gene_expression_cufflinks_gene` (
`gene_id` ,
`fpkm`,
`fpkm_percentile_compendium` ,
`fpkm_percentile_origin_tissue` ,
`fpkm_percentile_collection_tissue` ,
`fpkm_percentile_sample_cancer`
)
VALUES ('ENSG00000239906','555', NULL , NULL , NULL , NULL , NULL , NULL)

Eventually I want my trigger to insert the calculated NEW percentile values into the row being inserted, such as NEW.fkpm_percentile_compendium. Precalculating these percentile values is going to be necessary for performance reasons because this table will be > 500m rows, and eventually other percentile values will have to be computed based on relations, resulting in excessive joins.

I have a vague idea that the solution should involve SELECT INTO variables instead of a plain select statement, but this breaks my procedure.

DROP PROCEDURE IF EXISTS gecg_relative_expr_compendium |

CREATE PROCEDURE gecg_relative_expr_compendium( IN in_gene_id varchar(24), IN in_fpkm double, OUT percentile float(4,2) ) 
BEGIN 
    SET @row := 0;
    SELECT fpkm, ( (rank / @row) *100 ) AS percentile INTO fpkm, percentile
    FROM (
        SELECT fpkm, @row := @row +1 AS rank, gene_id
        FROM gene_expression_cufflinks_gene
        WHERE gene_id = in_gene_id
        ORDER BY fpkm ASC
    ) AS p
    WHERE fpkm = in_fpkm;
END |

gives me:

#1327 - Undeclared variable: fpkm 

So my questions are

1> How can I access the new vars I've selected INTO from within the same procedure?

2> how can I do this in such a way as to return the vars to the trigger without returning a result set?

Thanks so much for your help.

---

UPDATE >> Here are the working procedure and trigger:

DROP PROCEDURE IF EXISTS gecg_relative_expr_compendium |
CREATE PROCEDURE gecg_relative_expr_compendium( IN in_gene_id varchar(24), IN in_fpkm double, OUT out_fpkm double, OUT out_percentile float(5,2)  ) 
BEGIN 
    SET @row := 0;
    SELECT fpkm, ( (rank / @row) *100 ) AS percentile
    FROM (
        SELECT fpkm, @row := @row +1 AS rank, gene_id
        FROM gene_expression_cufflinks_gene
        WHERE gene_id = in_gene_id
        ORDER BY fpkm ASC
    ) AS p
    WHERE fpkm = in_fpkm
    LIMIT 1 
    INTO out_fpkm, out_percentile;
END |



DROP TRIGGER IF EXISTS insert_gecg_relative_expr |

CREATE TRIGGER insert_gecg_relative_expr

 BEFORE INSERT ON `gene_expression_cufflinks_gene` FOR EACH ROW  
 BEGIN  
    CALL gecg_relative_expr_compendium(NEW.gene_id, NEW.fpkm, @f, @p);
    SET NEW.fpkm_percentile_compendium = @p;
 END  
|

The trigger runs the percentile calculation on a row as it gets inserted. Now on to the next task, to run that procedure on all the other rows that match this gene_id. guhhh.

xoxoxox

Answer 1

Two things come to mind:

1) this is strange syntax:

SELECT fpkm, ( (rank / @row) *100 ) AS percentile INTO fpkm, percentile  

You can't use 'fpkm' until after this select statement.

2) its a bit hokey but one way to 'return' values from a procedure is to put them into a temp table. Then select them out from the calling process.

Answer 2

SELECT fpkm, ( (rank / @row) *100 ) AS percentile INTO fpkm, percentile
                                                       ^^^^

there's your undeclared var. probably should be in_fpkm or other name.