espertechinc / esper

Esper Complex Event Processing, Streaming SQL and Event Series Analysis

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

No output from statement with multiple regexp or like clauses for the same event property

frbor opened this issue · comments

We're having problems with a statement with multiple REGEXP clauses. To replicate the issue I've created an esper notebook with these cells:

%esperepl

CREATE SCHEMA Event(timestamp long, domain string);

SELECT event 

FROM Event(
        domain REGEXP "test.*"
    AND domain NOT REGEXP ".*\\.gov"
    AND domain NOT REGEXP ".*\\.org"
) event
%esperscenario

Event = {domain="test.com"};

When running this notebook I do not get any output events, but if I remove either one of the AND statements I get output.

I know that I can refactor this simple example using a single REGEXP, but I am puzzled why this breaks in the first place as I can not see any logical explanation.

We've tried the same code locally, and the statement works in Esper 8.4.0, but not in Esper 8.5.0 and higher.

We observe the same issue with the LIKE statement, do the example below does not match either (unless you remove one of statements:

SELECT event 
FROM Event(
    domain LIKE "%" AND domain LIKE "%"
) event

Again, this is a simplified example, but just shows that it seems to be a problem when matching with multiple LIKE or REGEXP?

We tried setting FilterIndexPlanning.BASIC and indeed, that seems to mitigate the problem.

Is there any updates on this issue? I managed to reproduce the same issue both locally and by the esper notebook example as @frbor provided.