To put it briefly, a program matches /rule*input/

A rule is composed of regexp, -> sign and something I call a segexp (just because S goes after R)

A regexp is a string enclosed in double quotes. *, + (instead of |) and parentheses work as you'd expect them. There is an additional element - a name prefix. This is a string enclosed in [] followed by a regexp. This name is going to be used by segexp. Name prefix has precedence between * and +. To use symbols ", *, +, (, ), [, ], \ in a regexp as normal characters, prefix them with a \.

Example: "[x](0+1)*1"
This matches all binary strings that end with 1. All characters but the last one are given a name [x].

A segexp is a string enclosed in double quotes. It is composed of symbols and names. A name is the same as in regexp. Whatever was assigned to it will be inserted here. If that name was not used, nothing is inserted (no error either). Symbols ", [, ], \ may need a \ prefix.

Example: "[x]0[x][y]"
If used with the regexp of last example and a string 101, it will produce a string 10010

An input is again a string in ". It only has characters. only " and \ need a \.

The input is matched to each rule in the order they are written in the file. If a match occurs, input is replaced with whatever segexp built and matching starts from the top of the list again.
When no rule matches, the current string is printed.