How Ordered Regex Parsers Decide a Match

The dominant design for user agent parsing is an ordered list of patterns evaluated until one matches. The order is not incidental; it carries most of the parser's knowledge.

First match wins by design

The parser walks its list from the top and stops at the first pattern that matches the string. Later patterns never run, whether or not they would also have matched.

This makes the list a decision procedure rather than a set of rules. Two patterns that both match express a priority relationship through their positions.

Evaluating all patterns and choosing the best match is possible but far slower, and defining "best" requires information the patterns do not carry.

Specific patterns must precede general ones

Strings are built by inheritance, so a niche browser's string usually contains the tokens of the mainstream browser it is based on. The mainstream pattern matches it too.

The niche pattern must therefore sit above the general one. Placed below, it never runs, and the browser is reported as whatever it inherited from.

This single rule explains most misidentification in the wild. The pattern exists, matches correctly in isolation, and is unreachable in its position.

Ordering encodes vendor relationships

Reading a well-maintained list top to bottom describes the inheritance structure of the browser ecosystem: derivatives above their bases, vendor forks above the shared engine.

That structure changes as products fork, rebase or are discontinued. A list that is correct today can misorder within a year without any pattern being wrong.

Maintenance is therefore mostly reordering rather than rewriting. New entries usually go near the top of their family, not at the end of the file.

Adding a pattern is riskier than it looks

Appending a pattern at the end is safe and usually useless, because something above it already matches. Inserting higher is effective and can capture strings intended for other entries.

A pattern loose enough to match its target may match a whole family, silently taking over entries below it. The symptom is a sudden shift in reported browser share.

Testing an insertion requires running the full corpus before and after and comparing outputs, not checking that the new string parses correctly.

Why the design survives its awkwardness

Ordered patterns are fast, easy to express as data, and simple enough for contributors to extend. Those properties matter more than elegance for a dataset updated continuously.

Alternative designs improve on accuracy or speed but demand more structure than volunteer contributions reliably supply. The ordered list absorbs messy input well.

Its weakness is that correctness is a property of the whole list rather than of any entry. That is worth remembering before editing one.