Variants

All

All capture states, including those corresponding to both implicit and explicit capture groups, are included in the Thompson NFA.

Implicit

Only capture states corresponding to implicit capture groups are included. Implicit capture groups appear in every pattern implicitly and correspond to the overall match of a pattern.

This is useful when one only cares about the overall match of a pattern. By excluding capture states from explicit capture groups, one might be able to reduce the memory usage of a multi-pattern regex substantially if it was otherwise written to have many explicit capture groups.

None

No capture states are compiled into the Thompson NFA.

This is useful when capture states are either not needed (for example, if one is only trying to build a DFA) or if they aren't supported (for example, a reverse NFA).

Warning

Callers must be exceedingly careful when using this option. In particular, not all regex engines support reporting match spans when using this option (for example, PikeVM or BoundedBacktracker).

Perhaps more confusingly, using this option with such an engine means that an is_match routine could report true when find reports None. This is generally not something that should happen, but the low level control provided by this crate makes it possible.

Similarly, any regex engines (like meta::Regex) should always return None from find routines when this option is used, even if some of its internal engines could find the match boundaries. This is because inputs from user data could influence engine selection, and thus influence whether a match is found or not. Indeed, meta::Regex::find will always return None when configured with this option.