README.md

PCOV

A self contained CodeCoverage compatible driver for PHP7

Requirements

• PHP 7.1+

API

/*
* Shall start recording coverage information
**/
function \pcov\start(void) : void;

/*
* Shall stop recording coverage information
**/
function \pcov\stop(void) : void;

/*
* Shall collect coverage information
* @param integer \pcov\all        shall collect coverage information for all files
*		 \pcov\inclusive  shall collect coverage information for the specified files
*		 \pcov\exclusive  shall collect coverage information for all but the specified files
* @param array   filenames
* Note: paths in filter must be realpath
*/
function \pcov\collect(int $type = \pcov\all, array $filter = []) : array;

/*
* Shall clear stored coverage information
* @param set true to clear file cache
* Note: clearing the file tables may have surprising consequences
*/
function \pcov\clear(bool $files = false) : void;

/*
* Shall return list of files waiting to be collected
*/
function \pcov\waiting() : array;

/*
* Shall return the current size of the trace and cfg arena
*/
function \pcov\memory() : int;

Configuration

PCOV is configured using PHP.ini:

Notes

The recommended defaults for production should be:

• pcov.enabled = 0

The recommended defaults for development should be:

• pcov.enabled = 1
• pcov.directory = /path/to/your/source/directory

When pcov.directory is left unset, pcov will attempt to find src, lib or, app in the current working directory, in that order; If none are found the current directory will be used, which may waste resources storing coverage information for the test suite.

If pcov.directory contains test code, it's recommended to set pcov.exclude to avoid wasting resources.

To avoid unnecessary allocation of additional arenas for traces and control flow graphs, pcov.initial.memory should be set according to the memory required by the test suite, which may be discovered with \pcov\memory().

To avoid reallocation of tables, pcov.initial.files should be set to a number higher than the number of files that will be loaded during testing, inclusive of test files.

Note that arenas are allocated in chunks: If the chunk size is set to 65536 and pcov require 65537 bytes, the system will allocate two chunks, each 65536 bytes. When setting arena space therefore, be generous in your estimates.

Interoperability

When pcov is enabled by configuration pcov.enabled=1:

• interoperability with Xdebug is not possible
• interoperability with phpdbg is not possible

At an internals level, the executor function is overriden by pcov, so any extension or SAPI which does the same will be broken.

When pcov is disabled by configuration pcov.enabled=0:

• pcov is zero cost - code runs at full speed
• Xdebug may be loaded
• phpdbg may be executed

At an internals level, the executor function is untouched, and pcov allocates no memory.

Differences in Reporting

There are subtle differences between Xdebug and PCOV in reporting: Both Xdebug and PCOV perform branch analysis in order to detect executable code. Xdebug has custom written (very mature, proven) analysis, while PCOV uses the very well proven control flow graph from Optimizer. They generate comparably accurate reports, while phpdbg uses less robust detection of executable code and generates reports with known defects. One such defect in phpdbg is this:

/* 2 */ function foo($bar) {
/* 3 */ 	if ($bar) {
/* 4 */			return true;
/* 5 */		}
/* 6 */	}

phpdbg will detect that this function is 100% covered when the first control path is taken, if ($bar), because it cannot correctly detect which implicit return paths inserted by Zend at compile time are executable, and so chooses to ignore them all. While this may seem like a trviail difference to some, it means that the reports generated by phpdbg are not completely trustworthy.

While the accuracy of Xdebug and PCOV are comparable, the reports they generate are not precisely the same, one such example is the switch construct:

/* 2 */ switch ($condition) {
/* 3 */		case 1:
/* 4 */			return "PHP7 rox!";
/* 5 */	}

From PHP 7.2.15 and PCOV 1.0, PCOV will detect the executability of the cases inside the switch body correctly, but will not detect line 2 (with the switch statement) as executable because Zend didn't output an executable opcode on that line. Xdebug's custom analysis doesn't use the same method and so will show an extra executable line on 2. Pre 7.2.15 and PCOV 1.0, the coverage of some switches is questionable as a result of the way Zend constructs the opcodes in the body of the switch - it may not execute them depending on a jump table optimization.

While Xdebug and PCOV both do the same kind of analysis of code, Xdebug is currently able to do more with that information than just generate accurate line coverage reports, it has path coverage and PCOV does not, although path coverage is not yet implemented (and probably won't be) by CodeCoverage.

Differences in Performance

The differences in performance of Xdebug and PCOV are not slight. Xdebug is a first and foremost a debugging extension, and when you load it, you incur the overhead of a debugger even when it's disabled. PCOV is less than 1000 lines of code (not including CFG) and doesn't have anything like the overhead of a debugger.