[Bro-Dev] [JIRA] (BIT-1072) merge topic/bernhard/hyperloglog

Robin Sommer (JIRA) jira at bro-tracker.atlassian.net
Fri Aug 30 11:30:00 PDT 2013


    [ https://bro-tracker.atlassian.net/browse/BIT-1072?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13819#comment-13819 ] 

Robin Sommer commented on BIT-1072:
-----------------------------------

I ended up refactoring and reformatting this quite a bit, it's in topic/robin/hyperlolog-merge. However, I broke something, the tests aren't working. Need to debug that later. In the meantime, some requests/questions:

- Please look over my changes and see if they make sense. (You don't need to track down the bug; I take the blame for that :).

- Can you please rework the Doxygen comments in HyperLogLog.h so that the descriptions for the public methods are understandable on their own. Right now I can't really follow them as often they talk about internal parameters/functionality. What you could do is provide a short overview of the data structure parameters in the class' doc string, and then refer to that in the methods. Also, please use the @param and @return syntax. (Start from my branch with this: I already reformatted and reordered things there quite a bit.)

- I don't understand what can be parameterized by the user and what not (and why not). One can give an error margin to the actor, but the confidence is a compile time constant. Also, where are the magic alpha_m values in *.cc coming from? Are these indeed always static values that don't depend on any parameters?


                
> merge topic/bernhard/hyperloglog
> --------------------------------
>
>                 Key: BIT-1072
>                 URL: https://bro-tracker.atlassian.net/browse/BIT-1072
>             Project: Bro Issue Tracker
>          Issue Type: New Feature
>          Components: Bro
>    Affects Versions: git/master
>            Reporter: Bernhard Amann
>             Fix For: 2.2
>
>         Attachments: out.pdf
>
>
> The branch adds support for the hyperloglog data structure.
> In the branch, core/leaks/basic-cluster.bro currently faisl. However, this seems to be unrelated to hll and just to be triggered by the addition of it to the sumstats tests. It looks like some kind of scriptland issue. pprof output is attached. (master, workers don't leak memory)



--
This message was sent by Atlassian JIRA
(v6.1-OD-06#6139)


More information about the bro-dev mailing list