10. History

Updates to Version 1.0.0

‘+’ and ‘*’ are now unlimited (ie not restricted to 65535 repetitions) Loop optimisation has now been extended to inner loops (previously just those at the start of an expression.). This improves performance significantly for expressions with embedded loops that are likely to capture long strings.

The matching sematics have now been clarified (thanks for several comments) so that the ‘first longest’ rule usually applies following POSIX, assuming that shortest (reluctant) matching is not specified for a group. Reluctant matching is still supported.

Bugfixes: The anchor movement process described below at beta 5 has been superceeded, since it was unsafe in the presence of invalid encodings that resulted from parsing inputs with sequences of different encodings. (‘.*’ does not now re-position the anchor after newline on failure.) This change does not impact performance.

Updates to Version 1.0.0 beta 7

Minor bugfix (match() not in __init__, purge function)

Updates to Version 1.0.0 beta 6

Indexing alternative expressions (INDEXALT flag) now applies to any alternative expression, not just to keywords.

Note

The property of the Match object which has the value of the matched pattern is keypattern. The original keyword property has been removed.

Match objects now have a flags property which are the flags used to compile the matched expression.

Updates to Version 1.0.0 beta 5

__getstate__ and __setstate__ functions have been added to the extension class (jsvm) to allow a compiled expression to be pickled. This is necessary for use with the multiprocessing module on Windows.

Backreferences are now supported.

match() and purge() functions have been added.

Reluctant (lazy) quantifiers (e.g. +?) are now supported in addition to the reluctant group extension; see Reluctant Matching for details.

The memory footprint is still limited below 10MB but its size is now adaptive depending on the expression to be evaluated.

The depreciated ASYNCHRONOUS flag has been removed.

There are minor changes to syntax for compatibility with other matching engines; the most significant is that the default status of repeated groups and groups that encompass the whole expression has changed from non-capturing to capturing.

All repeats are limited to a maximum of 65535. Note that if the expression starts with ‘’.*’’ or similar then this means that if a match fails then the next match will be attempted after the following newline. However if DOTALL is set the anchor will be incremented and a match will be attempted from every byte - the user can prevent this behaviour by setting an anchor stop position.

Updates to Version 1.0.0 beta 4

The primary update to beta 3 is to automatically detect expressions that start with .* and other similar prefixes. In previous versions it was necessary to set the anchor stop at the start of the text buffer to prevent the vm uselessly restarting a failed match from every position in the buffer. These cases are now detected automatically.

This modification has also improved the use of Multiline ^: the fast prefix scanning available in the VM is now used to move the anchor point to each line start.

The update also includes bug fixes (optimiser treatment of non-consuming groups and thread order processing in counted loops) which appear seldom in (my) practice but improve asyptotic performance.