Regular Expression Tutorial

Introduction

This section offers a real example of the XReplace-32 regular expressions mechanism. It explains step by step a complex replacement that saved hours of time to a real XReplace-32 user. This is a live problem from the industry that was impossible to solve without regular expressions.

Solving a Real Problem

There're hundreds of files of the following format:

Software - Package - (Path:C:\WINNT\SYSTEM32\NTOSKRNL.EXE) - Version = 4.0 Hardware - Package - (Path:T:\32bit\system\4nt\PRODUCTS.TXT) - Name = Microsoft Windows NT SP3 Software - Package - (Path:N:\32bit\Xreplace\xrep32.exe) - File Size = 914688 Hardware - Package - (Path:C:\WINNT\SYSTEM32\MFC42.DLL) - File Date = 862437600 No

The task is to remove paths inside the brackets to produce the following output:

Software - Package - NTOSKRNL.EXE - Version = 4.0 Hardware - Package - PRODUCTS.TXT - Name = Microsoft Windows NT SP3 Software - Package - xrep32.exe - File Size = 914688 Hardware - Package - MFC42.DLL - File Date = 862437600 No

Writing Regular Expression Ranges

A regular expression is a series of patterns that are matched to the real data. Once the pattern is matched, it's cut into pieces following the pattern format. Each piece is identified by a number and can be altered, copied or replaced.

A range defines what kind of characters a pattern can contain. For example, all capital letters range from A to Z. XReplace-32 regexp range for capital letters is written

[A-Z]

. A range for all alphanumeric characters is thus:

[A-Z,a-z,0-9]

. To include a single bracket ( into the pattern you'll have to use a backslash:

[A-Z,a-z,0-9,\(]

Writing Regular Expression Patterns

Patterns are separated in XReplace-32 by putting them into parenthesis. A full pattern of all alphanumeric characters is written

([A-Z,a-z,0-9])

and will match any character between A and Z, a and z or 0 and 9. To match a sequence of characters it is necessary to express the fact that a pattern is repeated multiple times (as many as possible). This is done by adding a * to the pattern:

([A-Z,a-z,0-9]*)

Multiple patterns are easy to write. For example, a filename is of the form name.extension and will be matched by

([A-Z,a-z,0-9]*\.[A-Z,a-z,0-9]*)

This reads: a sequence of alphanumeric characters followed by a dot and an another sequence of alphanumeric characters.

Getting to the Point

We can now attempt to get rid of (Path:?:\ and replace it by a simple backslash in order to obtain:

Software - Package - \WINNT\SYSTEM32\NTOSKRNL.EXE) - Version = 4.0

for the fisrt line.

The source replacement is:

(\(Path:)([A-Z,a-z]\:\\)

and the target is

\\

which will match a left parenthesis followed by Path: , a single character, a column and a backslash.

We must now get rid of all sequences between backslashes:

Software - Package - \WINNT\SYSTEM32\NTOSKRNL.EXE) - Version = 4.0 Software - Package - \SYSTEM32\NTOSKRNL.EXE) - Version = 4.0 Software - Package - \NTOSKRNL.EXE) - Version = 4.0

There's no mechanism to repeat a replacement till no more occurrences are found. We'll have to repeat the operation that will replace a backslash followed by a sequence of alphanumeric characters (directory names) and terminated by another backslash.

Rewriting Patterns

The replacement is almost finished. It's obviously easy to replace a filename starting by a backslash and ending with a right parenthesis. The source pattern sequence is:

(\\)([A-Z,a-z,0-9]*\.[A-Z,a-z,0-9]*)(\))

As we replaced by a simple backslash in previous steps, we'll have to rewrite the filename as the result of the replacement. Remember that each pattern could be identified by a number. In this example, patterns are the following:

1: \\ 2: [A-Z,a-z,0-9]*\. [A-Z,a-z,0-9]* 3: \)

The target replacement is \2 , which is the filename only.

Getting it Together

The full replacements sequence is finally:

Imagine making such a replacement in 10'000 files by hand!