HS_FLAG_DOTALL | HS_FLAG_SOM_LEFTMOST

[sadegh01]

These 2 flags are conflicted .
When I want to Use these two flags together the flag HS_FLAG_SOM_LEFTMOST does not work correctly and at the eventHandler() function the value of from is always 0 zero !!!

[jviiret]

These two flags should definitely work together. Please post a test case with the following information:

• The full expression, flags and mode passed to hs_compile(),
• The data being scanned,
• The locations of the matches you are receiving, and where you expect them to be.

[sadegh01]

These are my flag's
flags.push_back(HS_FLAG_DOTALL | HS_FLAG_SOM_LEFTMOST );

This is My pattern to extract javascript :
1:/<script[^/][^&gt;]>(.?)</script[^&gt;]>|<javascript[^/][^&gt;]>(.?)</javascript[^&gt;]>/

This is My File :
<script type="text/javascript">//<![CDATA[si_ST=new Date;//]]></script>
lablablab
<script type="text/javascript">//<![CDATA[
_G.HT=new Date;
//]]></script></html>
lablablab

I want to use this (HS_FLAG_DOTALL) to detect New Line and use this (HS_FLAG_SOM_LEFTMOST) to get start point .

Comment : I used Escape HTML to copy My file and My pattern here . for your test would you please use this link
http://www.freeformatter.com/html-escape.html#ad-output to change it to unscaped mode

[jviiret]

I think your escaped markup might have stripped some characters from your pattern - I'm assuming from your description that this is what it should look like: (on github, indenting with four spaces will make their Markdown support render text as code without formatting)

/<script[^/][^>]*>(.*?)</script[^>]*>|<javascript[^/][^>]*>(.*?)</javascript[^>]*>/

I think the issue here is that you are assuming backtracking semantics, whereas Hyperscan provides automata semantics. This means that instead of providing one "best match" that takes into account greedy/ungreedy repeats, alternation ordering, etc like PCRE, Hyperscan delivers all possible matches for a given regex. In these semantics, there is no difference between .* and .*?.

This is a fundamental difference from the way that a backtracking matcher like PCRE operates. We have a more detailed description of it in the Semantics section of the Hyperscan developer reference.

In this particular case, this is why SOM_LEFTMOST is always reporting a from offset of zero: your .*? repeats in DOTALL mode will match any sequence of characters, so the leftmost start of any match is the first occurrence a match for of <script[^/][^>]*>, which is the <script type="text/javascript"> at offset 0 of your file.

I would suggest that the easiest way to use Hyperscan to extract the data between two script tags would be to split your pattern up into four patterns:

1:/<script[^/][^>]*>/
2:/</script[^>]*>/
3:/<javascript[^/][^>]*>/
4:/</javascript[^>]*>/

You can then track the offsets at which patterns 1 and 2 match, extracting the data between them, and similarly for 3 and 4.

[sadegh01]

Dear jviiret
Thank you a lot for your attention and explanation ;)