Struct fst::Regex [−] [src]

pub struct Regex {
    // some fields omitted
}

A regular expression for searching FSTs with Unicode support.

Regular expressions are compiled down to a deterministic finite automaton that can efficiently search any finite state transducer. Notably, most regular expressions only need to explore a small portion of a finite state transducer without loading all of it into memory.

Syntax

Regex supports fully featured regular expressions. Namely, it supports all of the same constructs as the standard regex crate except for the following things:

  1. Lazy quantifiers, since a regular expression automaton only reports whether a key matches at all, and not its location. Namely, lazy quantifiers such as +? only modify the location of a match, but never change a non-match into a match or a match into a non-match.
  2. Word boundaries (i.e., \b). Because such things are hard to do in a deterministic finite automaton, but not impossible. As such, these may be allowed some day.
  3. Other zero width assertions like ^ and $. These are easier to support than word boundaries, but are still tricky and usually aren't as useful when searching dictionaries.

Otherwise, the full syntax of the regex crate is supported. This includes all Unicode support and relevant flags. (The U and m flags are no-ops because of (1) and (3) above, respectively.)

Matching semantics

A regular expression matches a key in a finite state transducer if and only if it matches from the start of a key all the way to end. Stated differently, every regular expression (re) is matched as if it were ^(re)$. This means that if you want to do a substring match, then you must use .*substring.*.

Caution: Starting a regular expression with .* means that it could potentially match any key in a finite state transducer. This implies that all keys could be visited, which could be slow. It is possible that this crate will grow facilities for detecting regular expressions that will scan a large portion of a transducer and optionally disallow them.

Example

This example shows how to run a regular expression on a Set.

use fst::{IntoStreamer, Streamer, Regex, Set};

let set = Set::from_iter(&["foo", "foo1", "foo2", "foo3", "foobar"])
              .unwrap();

let re = Regex::new("f[a-z]+3?").unwrap();
let mut stream = set.search(re).into_stream();

let mut keys = vec![];
while let Some(key) = stream.next() {
    keys.push(key.to_vec());
}
assert_eq!(keys, vec![
    "foo".as_bytes(), "foo3".as_bytes(), "foobar".as_bytes(),
]);

Warning: experimental

While executing a regular expression against a finite state transducer will be very fast, construction of a regular expression automaton may not be. Namely, this implementation is a proof of concept. In particular, one of its major deficiencies is that it can use enormous amounts of memory. Note though, that the construction phase will return an error if the underlying automata grows too big (tens of MB).

This is important functionality, so one should count on this implementation being vastly improved in the future.

Methods

impl Regex
[src]

fn new(re: &str) -> Result<Regex>

Create a new regular expression query.

The query finds all terms matching the regular expression.

If the regular expression is malformed or if it results in an automaton that is too big, then an error is returned.

A Regex value satisfies the Automaton trait, which means it can be used with the search method of any finite state transducer.

Trait Implementations

impl Automaton for Regex
[src]

type State = Option<usize>

The type of the state used in the automaton.

fn start(&self) -> Option<usize>

Returns a single start state for this automaton. Read more

fn is_match(&self, state: &Option<usize>) -> bool

Returns true if and only if state is a match state.

fn can_match(&self, state: &Option<usize>) -> bool

Returns true if and only if state can lead to a match in zero or more steps. Read more

fn accept(&self, state: &Option<usize>, byte: u8) -> Option<usize>

Return the next state given state and an input.

impl Debug for Regex
[src]

fn fmt(&self, f: &mut Formatter) -> Result

Formats the value using the given formatter.