PHP Oracle query select statement inside loop slow - php

Question

I have this php function to check and insert data from text file to database.
//Get All Model
$qModel = oci_parse($c1, "SELECT MODELID, MODEL_NAME FROM MEP_TBL_MODEL WHERE ACTIVE = 'Y' AND LOCATION = 'PCBA' ORDER BY MODELID ASC");
oci_execute($qModel);
while($dModel = oci_fetch_array($qModel))
{
//Configuration
$qDtl = oci_parse($c1, "SELECT * FROM MEP_TBL_MODEL_CONFIGURATION WHERE MODELID_FK = '" . $dModel['MODELID'] . "'");
oci_execute($qDtl);
while($dDtl = oci_fetch_array($qDtl))
{
$modelIDAccept[] = $dDtl['CONFIGURATIONID'];
$dateCode = date($dDtl['DATE_CODE']);
$readRowAfter = date($dDtl['READ_ROW_AFTER']);
$createFromFormat = $dDtl['CREATE_FROM_FORMAT'];
$ipAddress = $dDtl['IP_ADDRESS'];
$status = $dDtl['STATUS'];
if($dDtl['SOURCE'] != "")
{
$source = "\\".$dDtl['SOURCE'];
}
else
{
$source = "";
}
if(empty($ipAddress))
{
$fileAccept = file_get_contents("\\\\192.168.184.13\\Reports\\".$dModel['MODEL_NAME'].$source."\\Accept\\Accept_".$dDtl['MODEL_CODE']."_".$dateCode."_".$dDtl['TS_CODE'].".txt");
$linesAccept = explode("\n",$fileAccept);
$rowsintimespanAccept = 0;
for($i = $readRowAfter; $i < count($linesAccept); $i++)
{
$dateobjAccept = DateTime::createFromFormat($createFromFormat, $linesAccept[$i]);
if($dateobjAccept < $toDateTime && $dateobjAccept > $fromDateTime)
{
$rowsintimespanAccept++;
$logDate = $dateobjAccept->format('Y-m-d H:i:s');
//I put select query and insert here but it so slow.
$qChk = oci_parse($c1, "SELECT * FROM MEP_TBL_OUTPUT_DETAILS WHERE MODELID_FK = '" . $dModel['MODELID'] . "' AND RUNNING_DATE = TO_DATE('$logDate', 'YYYY-MM-DD hh24:mi:ss') AND TS_CODE = '" . $dDtl['TS_CODE'] . "' AND SHIFT = 'Morning' AND QUANTITY_STATUS = 'OK' AND CONFIGURATIONID_FK = '" . $dDtl['CONFIGURATIONID'] . "'");
oci_execute($qChk);
if(oci_fetch($qChk) > 0)
{
}
else
{
$qInsert = oci_parse($c1, "INSERT INTO MEP_TBL_OUTPUT_DETAILS(MODELID_FK, RUNNING_DATE, QUANTITY_STATUS, TS_CODE, SHIFT, CONFIGURATIONID_FK) VALUES('" . $dModel['MODELID'] . "', TO_DATE('$logDate', 'YYYY-MM-DD hh24:mi:ss'), 'OK', '" . $dDtl['TS_CODE'] . "', 'Morning', '" . $dDtl['CONFIGURATIONID'] . "')");
oci_execute($qInsert);
}
}
}
$totalAccept[] = $rowsintimespanAccept;
}
}
}
When I tried to run the code, I got very slow loading the page and sometimes it show me time out execution.
My question, is there any way to make the query fast maybe inside or outside the loop? I knew it slow because when I remove the select and insert query, the load page is only 3-4 seconds.

Answer 1

If I've read your code correctly, what you're after is a single MERGE statement that you can run on the database. I don't know PHP, so I can't give you how it should be called, but I can give you the SQL statement to run:
MERGE INTO mep_tbl_output_details tgt
USING (SELECT mtm.modelid,
mtm.model_name,
mtmc.configurationid,
mtmc.date_code,
mtmc.read_row_after,
mtmc.create_from_format,
mtmc.ip_address,
mtmc.status,
mtmc.ts_code
FROM mep_tbl_model mtm
INNER JOIN mep_tbl_model_configuration mtmc ON mtm.modelid = mtmc.modelid_fk
WHERE mtm.active = 'Y'
AND mtm.location = 'PCBA') src
ON (tgt.modelid_fk = src.modelid
AND tgt.ts_code = src.ts_code
AND tgt.configurationid_fk = src.configurationid
AND tgt.runningdate = :log_date
AND tgt.shift = 'Morning'
AND tgt.quantity_status = 'OK')
WHEN NOT MATCHED THEN
INSERT (tgt.modelid_fk, tgt.running_date, tgt.quantity_status, tgt.ts_code, tgt.shift, tgt.configuration_fk)
VALUES (src.modelid, :log_date, 'OK', src.ts_code, 'Morning', src.configurationid);
This does the join you were reinventing with your loops, links it back to the table you're trying to insert into, and only inserts a row if it doesn't already exist in the table.
You would need to write the PHP code to execute this, having passed the log_date in as a bind variable.
By binding the variable, you allow the database to skip the hard parse (i.e. finding out the best way to execute the query), which saves time.
By not fetching data and manually looping round before selecting more data and working out if you need to do the insert, you skip a whole lot of context switching and pulling/pushing data across the network. Let the database do the heavy lifting; it's what it's designed to do!