Regular expression based filters - uBlockOrigin/uBlock-issues GitHub Wiki
Adblock Plus filter syntax allows for the use of regular expression as filters.
It is not recommended to use regex-based filters, hence why I chose for a long time to not support these in uBlock: to prevent users who write their own filters from acquiring the bad habit of using regular expressions to filter network requests.
However this led to the often-repeated myth that the reason uBlock was efficient memory- and CPU-wise was because it did not support regex-based filters, a completely nonsensical assertion, given that there is a grand total of only 15 such filters in all of EasyList (at time of writing) out of tens of thousands.
It is because of this myth that I finally decided to support regex-based filters with version 0.8.6.0:
Given the way uBlock works internally, the regex-based filters are implemented in a more efficient way than other big-name blockers.
An efficient regex-based filter is one which does not need to be evaluated.
So if you decide to write a regex-based filter, here is the trick to help you make your regex as efficient as possible:
Use filter options to reduce the likelihood of a regex-based filter of being executed.
A most-efficient regex-based filter is one which comes with all the following filter options set:
- type: a filter which apply only to a specific request type will be executed only for request which matches the type.
- Example:
/\.filenuke\.com/.*\/[a-zA-Z0-9]{4}/$script
will be executed only for request of typescript
(this filter is found in EasyList)
- Example:
domain=
: The regular expression won't be executed if the hostname of a request does not match the hostnames declared in thedomain
filter optionthird-party
: the regular expression won't be executed if the request does not fulfill thethird-party
option (or it's complement~third-party
)
An example of a regex-based filter found in EasyList which is handled very efficiently by uBlock:
/http:.*(?:\+|\@|\=|\;|\_|\-|\!|\?|\&|\%|\#|\^|\:).*\/\//$script,third-party,domain=allenbwest.com
This filter contains all the filter options which makes it very unlikely that the regular expression will have to be executed. The regular expression will execute only if the request is of type script
, originates from allenbwest.com
, and is 3rd-party to allenbwest.com
.
If this sounds like basic common sense, it's because it is. However I've seen other big-name blockers out there execute all regex-based filters unconditionally for every request. (Edit: fixed in ABP)