$GLOBALS["PG_CONNECT"]=pg_connect(...);
....

function query ($sqlQuery,$var=0) {
   if (!$GLOBALS["PG_CONNECT"]) return 0;
   $lev=error_reporting (8); //NO WARRING!!
   $result=pg_query ($sqlQuery);
   error_reporting ($lev); //DEFAULT!!
   if (strlen ($r=pg_last_error ($GLOBALS["PG_CONNECT"]))) {
      if ($var) {
        echo "<p color=\"red\">ERROR:<pre>";
        echo $sqlQuery;
        echo "</pre>";
        echo $r;
        echo "&lt/p>";
      }
      close_db ();
      return 0;
   }
   return $result;
}

A quick note for novice users: when gathering input from fields on a web form that maintains a database connection, *never* use pg_query to do queries from the field. Always sanitize input using pg_prepare and pg_execute.

expanding on the note left by "cmoore" -

To check to see if the recordset returned no records,

<?php
  $result=pg_query($conn, "SELECT * FROM x WHERE a=b;");
  if  (!$result) {
    echo "query did not execute";
  }
  $rs = pg_fetch_assoc($result);
  if (!$rs) {
    echo "0 records"
  }
?>

-jack

It would be better this way:

<?php
  $result=pg_query($conn, "SELECT COUNT(*) AS rows FROM x WHERE a=b;");
  if  (!$result) {
   echo "query did not execute";
  }
  if ($line = pg_fetch_assoc($result)) {
    if ($line['rows'] == 0) {
     echo "0 records"
    }
  }
  else {
   while ($row = pg_fetch_array($result)) {
     //do stuff with $row
   }
  }
?> 

This solution doesn't raise the load of the system with the move of matching rows (perhaps 0,1, perhaps 100, 1000, ... rows)

One thing to note that wasn't obvious to me at first.  If your query returns zero rows, that is not a "failed" query.  So the following is wrong:
  $result=pg_query($conn, "SELECT * FROM x WHERE a=b;");
  if  (!$result) {
    echo "No a=b in x\n";
  }

pg_query returns FALSE if the query can not be executed for some reason.  If the query is executed but returns zero rows then you get back a resul with no rows.

There was a typo in the code that I posted:

<?php
  $result=pg_query($conn, "SELECT * FROM x WHERE a=b;");
  if  (!$result) {
   echo "query did not execute";
  }
  if (pg_num_rows($result) == 0) {
   echo "0 records"
  }
  else {
   while ($row = pg_fetch_array($result)) {
     //do stuff with $row
   }
  }
?>

Use pg_query to call your stored procedures, and use pg_fetch_result when getting a value (like a smallint as in this example) returned by your stored procedure.

<?php
$pgConnection = pg_connect("dbname=users user=me");

$userNameToCheckFor = "metal";

$result = pg_query($pgConnection, "SELECT howManyUsersHaveThisName('$userNameToCheckFor')");

$count = pg_fetch_result($result, 0, 'howManyUsersHaveThisName');
?>

Improving upon what jsuzuki said:

It's probably better to use pg_num_rows() to see if no rows were returned, as that leaves the resultset cursor pointed to the first row so you can use it in a loop.

Example:

<?php
  $result=pg_query($conn, "SELECT * FROM x WHERE a=b;");
  if  (!$result) {
   echo "query did not execute";
  }
  if (pg_num_rows($result) == 0) {
   echo "0 records"
  }
  else {
    while ($row = pg_fetch_array($result) {
      //do stuff with $row
    }
  }
?>

I, personally, also find it more readable.

Here is my small function to make it easier for me to use data from select queries (attention, it is sensitive to sql injection)
<?php
function requestToDB($connection,$request){ 
    if(!$result=pg_query($connection,$request)){
        return False;
    }
    $combined=array();
    while ($row = pg_fetch_assoc($result)) {
        $combined[]=$row;
    }
    return $combined;
}
?>

Example:
<?php
$conn = pg_pconnect("dbname=mydatabase");

$results=requestToDB($connect,"select * from mytable");

//You can now access a "cell" of your table like this:
$rownumber=0;
$columname="mycolumn";

$mycell=$results[$rownumber][$columname];
var_dump($mycell);