Hi, Firstly, the match order depends on the match offsets in your example. Patterns: 0: /12/ 1: /1/ 2: /13/ Input: 123 Since we find pattern 1 matches "1" at offset 1, pattern 0 matches "12" at offset 2, hyperscan will first trigger a match callback for pattern 1 at offset 1 and then trigger another one for pattern 0 at offset 2. Secondly, if there are patterns matching at the same offset, then the triggering order of match callbacks is arbitrary. i.e. Patterns 0: /[^c]*foobar/ 1: /[^h]{10, 20}foobar/ Input: badgebrushfoobartest If you define "HS_FLAG_SINGLEMATCH" for both pattern 0 and pattern 1, what hyperscan will do is to trigger match callbacks for pattern 0 and pattern 1 respectively at the first match offset, that is 16 in this case, and stop matching. Thanks, Xiang -----Original Message----- From: Hyperscan [mailto:hyperscan-bounces@lists.01.org] On Behalf Of Dmitry Kozlyuk Sent: Wednesday, April 18, 2018 10:14 PM To: hyperscan(a)lists.01.org Subject: [Hyperscan] Multi-pattern match order Hello, Could anyone clarify this part of the documentation: Multiple pattern matching: Hyperscan allows matches to be reported for several patterns simultaneously. This is not equivalent to separating the patterns by | in libpcre, which evaluates alternations left-to-right. Does it mean that if a text matches several patterns, the order of patterns reported to be matching will be completely arbitrary (i.e. depending on the whole pattern set), or there will be some predictable yet not left-to-right order, like from the least specific pattern to more specific ones? If it helps, I'm interested in the first match only. Flags are: HS_FLAG_DOTALL | HS_FLAG_ALLOWEMPTY | HS_FLAG_SINGLEMATCH. Example: patterns: "12", "1", "13" text: "123" matches: "1", "12" (least specific to most specific) -- Dmitry Kozlyuk

If 2 patterns match at the same offset like below example, then there will be 2 match callbacks happening in Hyperscan - one for pattern0 and the other one for pattern1. We don't deliver these callbacks inside Hyperscan in any specific order like least specific pattern to more specific ones you mentioned and can't predict or define the order by ourselves. Whether callback for pattern0 is delivered before or after that of pattern1 is mainly related to the database generated by hyperscan compiler. Thanks, Xiang -----Original Message----- From: Hyperscan [mailto:hyperscan-bounces@lists.01.org] On Behalf Of Dmitry Kozlyuk Sent: Thursday, April 19, 2018 8:06 PM To: Hyperscan regular expression matching library <hyperscan(a)lists.01.org&gt; Subject: Re: [Hyperscan] Multi-pattern match order Thank you, Xiang. However, I'm still confused with simultaneous "arbitrary" and "respectively" here: Could Hyperscan potentially first trigger callback for pattern 1, than for pattern 0 (still assuming a match at the same offset)? Empirically, it seems not to, but maybe I'm missing something in my tests, and also you've said "arbitrary". -- Dmitry Kozlyuk