README

Working with pivot tables

Example

Given the following input, e.g. from a logfile:

2015-05-10 00:00:09 OK: 304 100 bytes
2015-05-10 00:00:48 GOOD: 200 102 bytes (ETag: 214ceb4b-980-3a7bbd9630480)
2015-05-10 03:00:49 ERROR: 404 512 bytes (Not found)
2015-05-11 00:00:17 OK: 304 102 bytes
2015-05-11 02:01:01 ERROR: 500 0 bytes (Internal Server Error)
2015-05-11 02:01:02 ERROR: 500 256 bytes (Internal Server Error)
...

We will parse this using sscanf(), transforming the lines into arrays like the following:

["2015-05-10", "00:00:48", "GOOD", 200, 95, "ETag: 214ceb4b-980-3a7bbd9630480"]

We can the load this into our pivot table using the array offsets (if we had a map, we could use its string keys; for objects we'll pass references to the getters and for more complex situations we can pass closures). Putting it together, we get the following:

use io\streams\{TextReader, FileInputStream};
use util\data\PivotCreation;

$pivot= (new PivotCreation())
  ->groupingBy(2)        // category
  ->groupingBy(3)        // code
  ->spreadingBy(0)       // date
  ->summing(4, 'bytes')  // bytes
  ->create()
);

$reader= new TextReader(new FileInputStream('measures.log'));
while (null !== ($line= $reader->readLine())) {
  $pivot->add(sscanf($line, '%[0-9-] %[0-9:] %[^:]: %d %d bytes (%[^)])'));
}

The resulting table will look something like this (using "b:" as an abbreviation for bytes - this becomes relevant once we sum on multiple columns):

.------------------------------------------------------- ~ ---------------------------.
|                    || Columns                             |                         |
|                    ||--------------------------------- ~ -|                         |
| Category  | Count  || 2015-05-10    | 2015-05-11    |- ~ -| Sum        | Average    |
|-----------|--------||---------------|---------------|- ~ -|------------|------------|
| OK        | 2      || 1, b:100      | 1, b:102      |- ~ -| b:202      | b:101      |
| GOOD      | 1      || 1, b:102      |               |- ~ -| b:102      | b:102      |
| ERROR     | 3      || 2, b:512      | 1, b:256      |- ~ -| b:768      | b:256      |
| ^- client | ^- 1   || ^- 1, b:512   |               |- ~ -| ^- b:512   | ^- b:512   |
|   ^- 404  |   ^- 1 ||   ^- 1, b:512 |               |- ~ -|   ^- b:512 |   ^- b:512 |
| ^- server | ^- 2   || ^- 1, b:0     | ^- 1, b:256   |- ~ -| ^- b:256   | ^- b:128   |
|   ^- 500  |   ^- 2 ||   ^- 1, b:0   |   ^- 1, b:256 |- ~ -|   ^- b:256 |  ^- b:128  |
|-----------|--------||---------------|---------------|- ~ -|------------|------------|
| Total     | 6      || b:714         | b:358         |- ~ -| b:1072     | b:178.7    |
`------------------------------------------------------- ~ ---------------------------´

Accessing values in a pivot

The number of records grouped by the grouping columns can be retrieved via count(). The aggregates can be accessed by passing the category to the respective methods.

$count= $pivot->count('OK');                                // 2
$count= $pivot->count();                                    // 6

$count= $pivot->records('2015-05-10', 'OK');                // 1
$count= $pivot->records('2015-05-10');                      // 4

$transferred= $pivot->column('2015-05-10', 'OK')['bytes'];  // 100
$transferred= $pivot->column('2015-05-10')['bytes'];        // 714

$transferred= $pivot->sum('OK')['bytes'];                   // 202
$transferred= $pivot->sum()['bytes'];                       // 1072

$average= $pivot->average('OK')['bytes'];                   // 101.0
$average= $pivot->average()['bytes'];                       // 178.7

Drill down

We can dril down by the categories we grouped on by using the rows() method. To calculate the distribution of categories in percent of the total, we'll use the count() method.

$rows= $pivot->rows();                         // ['OK', 'GOOD', 'ERROR']

// OK: 2 / 6 = 33.3%
// GOOD: 1 / 6 = 16.7%
// ERROR: 3 / 6 = 50.0%
$total= $pivot->count();
foreach ($rows as $cat) {
  $count= $pivot->count($cat);
  printf("%s: %d / %d = %.1f%%\n", $cat, $count, $total, $count / $total * 100);
}

// client: 1
// server: 2
foreach ($pivot->rows('ERROR') as $code) {
  printf("ERROR %s: %dx\n", $row, $pivot->count('ERROR', $code));
}

It can also interesting to see a development over time, so we'll drill down based on the columsn instead.

$columns= $pivot->columns();                   // ['2015-05-10', '2015-05-11']

// 2015-05-10: 714 / 1072 bytes = 66.6%
// 2015-05-11: 358 / 1072 bytes = 33.4%
$total= $pivot->total()['bytes'];
foreach ($columns as $date) {
  $bytes= $pivot->column($date)['bytes'];
  printf("%s: %d / %d bytes = %.1f%%\n", $date, $bytes, $total, $bytes / $total * 100);
}