Path report from MySQL - php

Question

I have a database table that tracks all pageviews which are fired in via a php script. The table looks like this:
rowid (AI)
user_id
page_url
visitor_ip
session_id
I want to be able to query my table to "Show the paths (max 5 pages) visitors take to get to X page within a single session". The output would be a table with a URL in each column, so the path is left to right in the order they visited pages with the same session_id ending with a certain page.
Any clue? I've been looking for a reporting tool to help me build these segments but I'm not coming up with anything so I'm trying to see if there is a way to just query it. I'd like to avoid turning to some other tool for collection and just query my DB if I can.

Answer 1

Does something like this give you what you want (warning - untested):
select group_concat(page_url order by rowid separator '->'),session_id
from pageviews group by session_id
?

Answer 2

One idea is to use correlated subqueries in the select list.
If I understood the specification, the argument(parameter) to the query will be a specific `page_url`, given as "X" in the specification.
The outer query will retrieve the rows for that page_url. The subqueries in the SELECT list will get the previous page_url in the session. (We don't see a datetime/timestamp, so we will need to depend on values of `rowid` increasing for subsequent page views (i.e. previous page views will have a "lower" value of `row_id`.
Something like this:
SELECT ( SELECT p5.page_url
FROM pageviews p5
WHERE p5.session_id = t.session_id
AND p5.rowid < t.rowid
ORDER BY p5.rowid DESC
LIMIT 4,1
) AS back_5_page_url
, ( SELECT p4.page_url
FROM pageviews p4
WHERE p4.session_id = t.session_id
AND p4.rowid < t.rowid
ORDER BY p4.rowid DESC
LIMIT 3,1
) AS back_4_page_url
, ( SELECT p3.page_url
FROM pageviews p3
WHERE p3.session_id = t.session_id
AND p3.rowid < t.rowid
ORDER BY p3.rowid DESC
LIMIT 2,1
) AS back_3_page_url
, ( SELECT p2.page_url
FROM pageviews p2
WHERE p2.session_id = t.session_id
AND p2.rowid < t.rowid
ORDER BY p2.rowid DESC
LIMIT 1,1
) AS back_2_page_url
, ( SELECT p1.page_url
FROM pageviews p5
WHERE p1.session_id = t.session_id
AND p1.rowid < t.rowid
ORDER BY p1.rowid DESC
LIMIT 0,1
) AS back_1_page_url
, t.page_url
, t.session_id
, t.row_id
FROM pageviews t
WHERE t.page_url = 'X'
Those subqueries are going to be executed for each row returned by the outer query, so this could eat our lunch in terms of performance. If no suitable indexes are available, it's going to eat our lunch box too.
For the subqueries, we are going to want an index...
ON pageviews (session_id, row_id, page_url)
The outer query will benefit from an index ...
ON pageviews (page_url, row_id, session_id)
As an idea for the start of a different approach, if we were getting the path to every page_url, and not just a specific one...
SET group_concat_max_len = 524288 ;
SELECT t.session_id
, t.page_url
, SUBSTRING_INDEX(
GROUP_CONCAT(t.page_url SEPARATOR '\t' ORDER BY t.rowid DESC)
,'\t',6) AS `last_5_pages`
FROM pageviews t
GROUP
BY t.session_id
, t.page_url
HAVING t.page_url = 'X'
This assumes that the page_url will not contain a tab (0x09) character.
The last_5_pages column would be a tab-delimited list of page_url, the most recent page view first, followed by the previously viewed page_url, etc.
Getting those split out as individual columns would be more work, wrapping that query in an inline view and some combination of SUBSTRING_INDEX, possibly REVERSE, and a function to count the number of page_url in the list... that gets kind of nasty to do in SQL. If I went this approach, I'd prefer to handle parsing out the page_url from the tab delimited list in the client.

Answer 3

Here is what I ended up doing - worked great.
<?php
require_once 'init.php';
// ----------------- PAGE PATH REPORT
$html = "<table>";
$html .= "<tr><th align='left'>PAGE PATHS HITTING GOAL.PHP</th></tr>";
$paths = array();
$sql = "SELECT cookie_uid, page_url FROM pageviews ORDER BY rowid";
$result = mysqli_query($conn, $sql);
$got_rows = mysqli_num_rows($result);
if ($got_rows) {
while ($row = mysqli_fetch_array($result)) {
// Create an array for the cookie_uid if it doesn't exist yet
if ( ! array_key_exists($row['cookie_uid'], $paths) || ! is_array($paths[$row['cookie_uid']])) {
$paths[$row['cookie_uid']] = [];
}
// Add to the array now that we know it exists
array_push($paths[$row['cookie_uid']], $row['page_url']);
}
foreach ($paths as $session => $page) {
$html .= "<tr>";
$html .= '<td>' . implode(' ---> ', $page) . "</td>";
$html .= "</tr>";
}
} else {
$html .= '<td colspan="2">No results</td>' . "";
}
$html .= "</table>";
echo $html;
if (!mysqli_query($conn,$sql)) {
die('Error: ' . mysqli_error($conn));
}
// ----------------- ALL PAGES REPORT
echo "</br></br>";
echo "<tbody><table>";
echo "<tr><th align='left'>UNIQUE PAGES</th></tr>";
$sql = "SELECT distinct page_url FROM pageviews";
$allpages = mysqli_query($conn, $sql);
foreach ($allpages as $page){
echo "<tr>";
echo "<td>" . $page['page_url'] . "</td>";
echo "</tr>";
}
echo "</tbody></table>";
mysqli_close($conn);
error_reporting(E_ALL);
?>
That gives me this:
/analytics/testpage.php ---> /analytics/testpage2.php ---> /analytics/goal.php