Philip Collins


CGI::OptimalQuery - dynamic SQL query viewer


  use CGI::OptimalQuery;

  # construct a CGI::OptimalQuery object
  my $oq = CGI::OptimalQuery->new({
    q => CGI->new(),
    dbh => $dbh,

    title => 'Employee List',

    select => {
      'NAME' => ['emp','emp.lname||', '||emp.fname", "Name"],
      'DEPT' => ["dept", "", "Department"]

    # default shown columns alias list
    show => ['NAME', 'DEPT'],

    # defined joins included in the from clause
    joins => {
        'emp' => [ undef, 'employee', undef],
        'dept'=> ["emp", "
            left join ( SELECT name FROM dept ) dept
              on (emp.dept ="]
    options => {
      'CGI::OptimalQuery::InteractiveQuery' => {
        mutateRecord => sub { my $rec = shift; $$rec{NAME} = ''; }
        OQdataRCol => sub { my $rec = shift; return "stuff"; }

  # get the count

  # get link back to this report

  # output view to STDOUT


Developer describes environment, output options, and database query; CGI::OptimalQuery provides user with a web interface to view, filter, sort, and export the data.

Sounds simple, but CGI::OptimalQuery does not write the SQL for the developer. It is only responsible for gluing the appropriate pieces of SQL together to form an optimized SQL query, and outputing the results in a format the user chooses.

This module has been tested for:

  1) SQLite
  2) MySQL
  3) PostgreSQL
  4) Microsoft SQL Server (must use MARS_Connection=yes in connection option)
  5) Oracle

The important elements the developer describes are what fields (select elements) the user can see and what data sets (join elements) those fields come from. Each select and join element can depend upon joins. For every Optimal Query there is one driving data set. The driving set does not depend on other data sets. For every row in the driving data set there can only be one corresponding row when joining the driving data set to other joins described in the joins configuration hash reference. This allows Optimal Query to optimize SQL and only include the appropriate joins when the user has selected a column from one of those joins. For example: If there are employee and department tables and the user only wants to see employee fields (not department fields) then Optimal Query will not join in the department table.


Configure your web server to allow it to serve the "Resources" directory containing static html/js/css files. This directory is located near the installation path. The easiest way to find it is to execute:

  perl -MCGI::OptimalQuery -e '$_=$INC{"CGI/"};s/\.pm$/\/Resources\n/;print $_'

If you are using Apache, add the following config:

# required CGI::OptimalQuery static files Alias /OptimalQuery/ /usr/local/share/perl5/CGI/OptimalQuery/Resources/


new ( \%CONFIG )

Returns an optimal query object.

\%CONFIG (a hash reference) describes the environment, query description, and output options using key and value pairs. Possible configuration elements are shown below. (required ones are first)


The following KEY/VALUES below for %CONFIG in the call to new are required.

title => "This is the title of this Query"
dbh => DBI->connect( .. )

provides OptimalQuery a connection to the database.

show => ['COLALIAS1', 'COLALIAS2']

Default fields to show user when loading OptimalQuery the first time. This can also be set as a CGI param where the value is a comma separated list of column aliases. Example: "[NAME], [DOB]".

params => { show => "COLALIAS1,COLALIAS2", filter => "[COLALIAS1]=5", sort => "[COLALIAS1]", queryDescr => "descr here", hiddenFilter => "[COLALIAS1]=5" }

Get user set parameters from this hash instead of CGI params. Useful if you are constructing an OQ that should not be influenced by CGI params.

select => { SELECT_ALIAS => [ DEP, SQL, NAME, OPTIONS ], .. }

The select configuration describes what fields from the query can be selected, filtered, and sorted.


is the alias for the select field. This alias is used throughout the rest of the configuration to describe the field.


describes required joins that must be included to use the select specified. The DEP can be written as a string or an array reference of strings if multiple dependancies for the field exist.


SQL to display values for this field. Specified as a string or array reference where the first element is the SQL and each element after is a bind value.

NAME (STRING | undef)

label describing the field name. If undef, field cannot be selected by user and is considered hidden.


The following KEY/VALUES below describe OPTIONS used by the select configuration.

select => [COLALIAS1, COLALIAS2]
select => "COLALIAS1, COLALIAS2"

Define other select fields to be included in executed SQL. These fields can be used in custom formatters including the built in CGI::OptimalQuery::Base::recview_formatter.

formatter => CODEREF

sub { my ($val, $rec, $o, $colAlias) = @_; return $val; }

# Built in formatters to display all field/values specified in 'select' as text. \&CGI::OptimalQuery::Base::recview_formatter

html_formatter => CODEREF

sub { my ($val, $rec, $o, $colAlias) = @_; return $val; }

# Built in formatters to display all field/values specified in 'select' as html. \&CGI::OptimalQuery::Base::recview_html_formatter

db_formatter => CODEREF

sub { my ($val) = @_; return $val; }

# format a value before it is executed in the database

is_hidden => 1

hides the select field and data from being viewed by the user. Data for this select is still available in callbacks and can be included in the hiddenFilter.

always_select => 1

tells OptimalQuery to always select the column regardless if it isn't explicitly being used in the show. This does not automatically make it shown to the user, but it will be available to the developer in callbacks.

select_sql => (STRING | ARRAYREF)
filter_sql => (STRING | ARRAYREF)
sort_sql => (STRING | ARRAYREF)

SQL to use instead of the default SQL for the select for the context described.

date_format => (STRING)

if column is a date and date format is specified, OptimalQuery will write SQL to use the date format appropriately.

Note: Oracle's date component also has a built-in time component. If the data is '11/24/2005 14:56:45' and the date_format is 'MM/DD/YYYY', the date will show up as '11/24/2005'. If a user tries to filter on date '11/24/2005' Oracle will only match '11/25/2005 00:00:00' leaving out results the user probably thinks should be included. In this case, the developer should trunc the date. Trunc strips the time component from a date field. Example:

  DATE_COL => ['DEP1', 'trunc(dep1.date_field)', 'My Date',
                 { date_format => 'MM/DD/YYYY' } ]
joins => { JOIN_ALIAS => [ DEP, JOIN_SQL, WHERE_SQL, OPTIONS ], .. }

describes what tables to join in order to fulfill the dependancies used by the fields described in the SELECT HASHREF.


is the alias for the table or inline view decribed in the JOIN_SQL.


describes required joins that this join depends upon. This should be undef if and only if this is defining the driving data set.


describes the SQL that is used in the join clause for the generated SQL. Example: "LEFT JOIN dept ON (emp.dep_id =". If this describes the driving table, only the table name is needed. Inline views can also be used. Make sure you specify the alias on the view! Example: JOIN ( SELECT * FROM emp WHERE is_active = 'Y') active_emps


This is deprecated. It was used to describe the SQL in the where clause that was needed to join the table described in the from clause. Since SQL-92 allows developers to put the join SQL in the join, this should not be used.


The following KEY/VALUES below describe OPTIONS used by the joins configuration.

always_join => 1

tells OptimalQuery to always include join in query. Usfual when the join itself influences the number of results returned. Alternatively, an inline view could be constructed that performs the joins as part of the driving data set.

new_cursor => 1

tells OptimalQuery to open a new cursor for this join. This can be used to select and filter multi-value fields. Optionally, an order_by param can be specified to sort the results returned by the cursor as such:

new_cursor_order_by => ""

The following KEY/VALUES below for %CONFIG in the call to new are NOT required.

AutoSetLongReadLen => 1

Tells OptimalQuery to automatically set $dbh->{SetLongReadLen}. Used only in Oracle. Enabling this setting may slow down OptimalQuery since it needs to do extra queries to set the length if LOBS exist. This is only enabled by default when using Oracle.

check => 0

Tells OptimalQuery to do additional checking to make sure the amount of rows in the driving table is equal even when including other joins. It is off by default because there can be a significant performace hit when enabled.

debug => 0

sends debug info to the error_handler (STDERR is default)

error_handler => sub { ($err) = @_; }

intercept messages sent to the error handler. Very useful if you are running in a mod_perl env and want to redirect error messages using $areq->log_error($msg).

filter => "[SELECT_COL_ALIAS] like 'foo' AND .."
hiddenFilter => "[SELECT_COL_ALIAS] like 'foo' AND .."
forceFilter => "[SELECT_COL_ALIAS] like 'foo' AND .."

Filters add SQL to the where clause. If a CGI param called 'filter' or 'hiddenFilter' are provided, the CGI param value is used instead. The value of a forceFilter cannot be overridden. Users can manipulate a filter using the filter dialog tool. The hiddenFilter can only be manipulated using the GET param. Filters allow developers and end users to add SQL to the where clause. The filter grammar is described in the FILTER GRAMMAR section.

For example:

  <a href=/Search?filter=".escape_uri("[NAME] like 'foo'")
module => { OverloadModuleLabel => PerlModuleName, .. }

This is an advanced feature that can help perl guru's change the factory blueprints for optimal query modules instantiated by CGI::OptimalQuery.


allow developers to create complex predefined sql for insertion in the where clause by the 'filter' and 'hiddenFilter' parameters. There are two types of named_filters: "normal" and "custom". Normal named filters are defined with static SQL. Custom named filters are dynamic and most often take arguments which influence the SQL and bind params generated via callbacks.

  filterNameAlias => [ DEP, SQL, NAME ]

DEP is a string or an ARRAY of strings describing the dependancies used by the named filter. SQL is a string or an arrayref with SQL/bind values that is used in the where clause when the named filter is enabled. The NAME is used to describe the named filter to the user running the report.

  filterNameAlias => {
    title => "text displayed on interactive filter",

    html_generator =>
      sub { my ($q, $prefix) = @_; return $html;},

    sql_generator  => sub {
      my %args = @_;
      return [$deps, $sql, $name];

The html_generator is used by InteractiveFilter to collect input from the user. The sql_generator converts the named filter & arguments into deps, sql, and a name. The deps can be returned as an array ref of string deps if more than one dep exists. The sql can also be returned as an array ref where the first element is the sql and the rest are bind values.

named_sorts => { SortName => [ DEP, SQL, NAME], .. }

Named sorts aren't really used that often. They are really implemented for completeness and work the same way as named_filters.

options => { MODULENAME => { OPT_KEY => OPT_VAL, .. } }

OptimalQuery is made up of several modules. The 'options' configuration allows developers to configure these modules. See section InteractiveQuery Options.

  options => { 'CGI::InteractiveQuery' => \%opts }
output_handler => sub { print @_; }

override default output handler (print to STDOUT), by defining this callback.

q => new CGI()

Pass OptimalQuery thr CGI query object. OptimalQuery will automatically create a new CGI object if one is not passed in.

queryDescr => "Some text describing the query"

The query description is extra text describing the query and does not affect generated SQL. Can also be set as a CGI param.

resourceURI => "/OptimalQuery"

Path to optional OptimalQuery resources. Default path is shown.

results_per_page_picker_nums => [10,20,50,100,'All']

An interactive query displays a pager mechanism when the result set is larger than the rows_page param. This array reference allows a developer to override the default options a user can pick from the pager.

rows_page => 10

The default rows_page a user has when initially loading InteractiveQuery. Can also be set as a CGI param.

mutateRecord => sub { }
  $$schema{mutateRecord} = sub {
    my $rec = shift;

    # add html links to the person record 
    # if user selected the NAME field
    if (exists $$rec{NAME}) {
      $$rec{NAME} = "<A HREF=/PersonRecord?id=$$rec{ID}>".
canSaveDefaultSearches => 0 | 1

Indicate if user is allowed to set the default search using a saved search. This key/value must not be defined if default saved searches is not installed. To install this feature, add the is_default column to your oq_saved_search table.

savedSearchUserID => $user_id

InteractiveQuery can optionally save searches to a database so users can revisit them latter. To do this, saved searches are tied to a unique user id. See the "Saved Searches" section for more information on this topic.

savedSearchAlerts => 0 | 1

If savedSearchUserID, is defined, setting savedSearchAlerts to 1 will enhance dialogs to optionally allow users to configure saved searches to alert them when records are added, removed, or are present. See the "Saved Search Alerts" section for additional details on configuring this feature.

savedSearchAlertMaxRecs => 1000

Specify maximum record UIDs to store in saved search alert previously seen uid history. The previously stored list of UIDs is used to determine if a new record has appeared. Don't set this too high - the pks are stored in the saved search table. The default settings allow about 14k of data per saved search. (15 char key size * 1000 savedSearchAlertMaxRecs) / 1024 = 14k.

savedSearchAlertEmailCharLimit => 500000

Maximum character limit of saved saerch alert email. If limit is reached, a message appears instructing the user reduce the size by hiding fields or modify filters. The default value is 500000 (~ .5MB)


Default sort to show user when loading OptimalQuery for the first time. See the "SORT GRAMMAR" section for more information. Sort can also be specified as a CGI param.

state_params => [ 'form_field1', .. ]

If HTTP GET/POST params are required to dynamically generate a %CONFIG, the developer can specify the names of the GET/POST params in this array and OptimalQuery will automatically carry their state.

URI => "/URI/Back/To/This/Page"

The URI back to the page the user is currently on. The default URI is taken from the REQUEST_URI ENV.

URI_standalone => "/URI/Back/To/This/Page?layout=off"

By request, some developers use a separate URI to turn their layout system off so OptimalQuery can send the headers for content that can't be embedded.


Output view to output_handler (STDOUT is default).


Options for an InteractiveQuery can be set by defining the following HASHREF in %CONFIG.

  $CONFIG{options}{'CGI::OptimalQuery::InteractiveQuery'} = 
    { KEY => VAL, .. };
                - OR - 
  $CONFIG{options}{'CGI::OptimalQuery::InteractiveQuery'}{KEY} = {VAL};
disable_select => 1
disable_filter => 1
disable_sort => 1

Disables options for all fields. If you set all three your Optimal Query will work like a Pager.


Append additional HTML in the head section.

  # turns off default inline CSS
  $CONFIG{options}{'CGI::OptimalQuery::InteractiveQuery'}{htmlExtraHead} .=
    "<link href='/css/default.css' rel='stylesheet' type='text/css'>";
  # override the built-in edit link builder
  buildEditLink => sub {
    my ($o,$rec,$opts) = @_;
    return "/link/record?id=$$rec{U_ID};act=edit"'
  # override the built-in edit link builder
  buildNewLink => sub {
    my ($o,$opts) = @_;
    return "/link/record?act=new"'
color => '#cccccc'

specify the background color of the optimal query GUI.

useAjax => 1

Reload the data using ajax. Defaults to 1 unless not specified and usePopups is set to 0.

NewButton => "<a class=OQnewBut>new</a>"
editButtonLabel => 'edit'

OptimalQuery will automatically create an edit and new button if this is defined. When creating the link, OptimalQuery appends "?id=$$rec{U_ID};act=new;on_update=OQrefresh" or "?id=$$rec{U_ID};act=load;on_update=OQrefresh" to the link so the record module will know which view to load. "OQrefresh" is a function defined by Optimal Query that an external record module can call to update the Optimal Query window if a record has been updated.

htmlFooter => "<h1>this is a footer</h1>"
htmlHeader => "<h1>this is a header</h1>"
httpHeader => CGI::header('text/html')

override httpHeader content. If you prefer to not have InteractiveQuery send the header, set this value to empty string.

mutateRecord => sub { }
  mutateRecord => sub {
    my $rec = shift;

    # add html links to the person record 
    # if user selected the NAME field
    if (exists $$rec{NAME}) {
      $$rec{NAME} = "<A HREF=/PersonRecord?id=$$rec{ID}>".
noEscapeCol => ['NAME']

if certain columns should not be HTML escaped, let OptimalQuery know by adding them to this array.

OQdataLCol => sub { }
OQdataRCol => sub { }

Specify custom code to print the first or last column element. This is most often used to generate an view/edit button. If these callbacks are used, the editLink, and buildEditLink are ignored.

  OQdataLCol => sub {
    my ($rec) = @_; 
    return "<button onclick=\"OQopwin('/ViewRecord?id=$$rec{U_ID};on_update=OQrefresh';\">".
OQdocBottom => "bottom of the document (outside form)"
OQdocTop => "top of the doc (outside form)"
OQformBottom => "bottom of form"
OQformTop => "top of form"
usePopups => 1|0

use popups when opening a record form using the built in buttons The default is 1. If this is set to 1, useAjax default is 1.

WindowHeight => INT
WindowWidth => INT

Specify popup window width, height.

OQscript => " some javascript code (see examples below) "

OQscript gives you unlimited power to alter the output of optimal query by allowing you to enter javascript code that is executed client side For example to add a new command button:

  OQscript => "
var e = document.getElementById('OQcmds');
e.innerHTML = '".CGI::OptimalQuery::escape_js(q|<button type=button onclick="window.alert('hello there');">hello</button>|)."' + e.innerHTML; 


InteractiveQuery can optionally save searches to a database so users can revisit them latter. To do this, saved searches are tied to a unique user id. Developers should tell OptimalQuery the user id by defining 'savedSearchUserID' in their %CONFIG.

 $config{savedSearchUserID} = $user_id;

Saved Searches are stored in a table in the database pointed to by the database handle defined in $config{dbh}. The following table must exist before using saved searches.

  -- For mysql:
  CREATE TABLE oq_saved_search (
    user_id INT UNSIGNED,
      FOREIGN KEY fk_oqsavedsearch_userid (user_id)
    uri VARCHAR(100) NOT NULL,
    oq_title VARCHAR(1000) NOT NULL,
    user_title VARCHAR(1000) NOT NULL,
    params TEXT,
    -- 0:disabled, 1:added, 2:removed, 4:present
    alert_interval_min INT UNSIGNED,
    alert_dow VARCHAR(7),
    alert_start_hour INT UNSIGNED,
    alert_end_hour INT UNSIGNED,
    alert_last_dt DATETIME,
    alert_err TEXT,
    alert_uids LONGTEXT,
    is_default TINYINT(1) NOT NULL DEFAULT 0,
    CONSTRAINT unq_oq_saved_search UNIQUE (user_id,uri,oq_title,user_title)

  -- For Oracle:
  CREATE TABLE oq_saved_search (
    uri VARCHAR2(100) NOT NULL,
    oq_title VARCHAR2(1000) NOT NULL,
    user_title VARCHAR2(1000) NOT NULL,
    params CLOB,
    -- 0:disabled, 1:added, 2:removed, 4:present
    alert_mask NUMBER DEFAULT 0 NOT NULL,
    alert_interval_min NUMBER,
    alert_dow VARCHAR2(7),
    alert_start_hour NUMBER,
    alert_end_hour NUMBER,
    alert_last_dt DATE,
    alert_err CLOB,
    alert_uids CLOB,
    is_default NUMBER DEFAULT 0 NOT NULL,
    CONSTRAINT unq_oq_saved_search UNIQUE (user_id,uri,oq_title,user_title)
  CREATE SEQUENCE s_oq_saved_search;

Replace XYZ(id) with the name of the user table and primary key id.

Optimal Query also provides a canned "Show My Saved Searches" HTML form component that can be embedded inside a page. It can be used in the following manner:

  use use CGI::OptimalQuery();

  my $saved_searches_html = CGI::OptimalQuery::get_saved_search_list(
    $cgi_query, $dbh, $userid);
  print $saved_searches_html;

Use CSS to stylize the output.


Saved search alerts enhance Optimal Query to optionally allow users to configure saved searches to alert them when records are added, removed, or are present. To use this feature, follow the instructions in the "SAVED SEARCHES" section. You also need to make sure you set savedSearchAlerts to 1 in your schema if the current user is allowed to configure alerts.

Lastly, you need to create a cron that will execute your OptimalQuery.



    use strict;
    use FindBin qw($Bin);
    use lib "$Bin/../lib"; # include project lib
    use DBI();
    use CGI::OptimalQuery::SaveSearchTool();

    # connect to your database
    $APP::dbh ||= DBI->connect("dbi:SQLite:dbname=db/dat.db","","", { RaiseError => 1, PrintError => 1 });

      # default is shown
      # error_handler => sub { print STDERR @_; },

      # if debug is true, no email is sent, emails will be logged to the error_handler
      debug => 1,

      # database handle
      dbh => $APP::dbh,

      base_url => "",

      email_from => '',

      # define a handler which is called for each possible alert
      # alerts aren't actually sent until the very end where they are batched
      # and one email is sent for each email address containing one or more alerts
      handler => sub {
        # $o contains all the fields defined in the oq_saved_search rec
        my ($o) = @_;

        # you must set the email address for the search search owner id: $$o{USER_ID} 
        $$o{email_to} = '';

        # configure your application ENV
        # local $APP::q = new CGI();
        # local $APP::usr_id = $$o{USER_ID};
        # local %APP::data = ();

        # helper function to execute a perl module handler (will dynamically load module string and execute handler function)
        #if ($$o{URI} =~ /(\w+)\.pm$/) {
        #  CGI::OptimalQuery::SaveSearchTool::execute_handler("App::Applet::$1");

        # helper function to execute a perl script (will auto compile and cache a function)
        if ($$o{URI} =~ /(\w+\.pl)$/) {


  start: exp /^$/

     '(' exp ')' logicOp exp
   | '(' exp ')'
   | comparisonExp logicOp exp
   | comparisonExp

   | colAlias compOp colAlias
   | colAlias compOp bindVal

  bindVal: float | quotedString

   | /or/i

  namedFilter: /\w+/ '(' namedFilterArg(s? /,/) ')'

  namedFilterArg: quotedString | float | unquotedIdentifier

  unquotedIdentifier: /\w+/

  colAlias: '[' /\w+/ ']'

   | /\-?\d+\.?\d*/

   | /".*?(?<!\\)"/

     '<=' | '>=' | '=' | '!=' | '<' | '>' |
     /contains/i | /not\ contains/i | /like/i | /not\ like/i


  start: expList /^$/

  expList: expression(s? /,/)

     namedSort opt(?)
   | colAlias  opt(?)

  opt: /desc/i

  namedSort: /\w+/ '(' namedSortArg(s? /,/) ')'
  namedSortArg: quotedString | float

  colAlias: '[' /\w+/ ']'

   | /\-?\d+\.?\d*/

   | /".*?(?<!\\)"/


    Philip Collins
    University of New Hampshire



This program is free software licensed under the...

        The MIT License

The full text of the license can be found in the LICENSE file included with this module.