[ 
https://issues.apache.org/jira/browse/DRILL-4963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15834526#comment-15834526
 ] 

ASF GitHub Bot commented on DRILL-4963:
---------------------------------------

Github user arina-ielchiieva commented on the issue:

    https://github.com/apache/drill/pull/701
  
    @paul-rogers as we discussed to have tried to find the way to preserve 
lazy-init approach.
    I have renamed PR to reflect latest changes. Please find new solution and 
description of changes below:
     
    Lazy-init was performed only when function was not found during Calcite 
parsing but DRILL-4963 shows different cases when Calcite parsing can pass 
(usually during function overloading) but still function is not found. To 
handle such cases, we need to enhance lazy-init process:
    1. Lazy-init process should be more light-weight. Currently when function 
is not found, we load all jars from remote function registry and compare with 
jars from local function registry. It's not optimal especially when both 
registries are in sync. To improve performance I have introduced remove 
function registry version which can be used to check if we need to sync remote 
and local registries prior to checking jars and functions.
    2. During parsing stage we were only catching Calcite parsing exception but 
function not found can be also indicated by Drill function error and so on. Not 
to be engaged into enumerating possible exceptions (which can added in the code 
later), we are checking if remote and local function registries are in sync on 
any error. Such check may only affect on queries that will fail anyway. So if 
failure time will take a little bit longer, it won't make significant 
difference.
    3. During execution stage Drill attempts to find matching function from the 
list of functions with the same names. `DefaultFunctionResolver.getBestMatch()` 
does not do exact match, it may return function with different input parameters 
types. Best match is found according to rules described in 
`TypeCastRules.class`. Currently we attempt to sync remote and local function 
registries only if best matching function was found but it is not correct since 
even if Drill finds the best matching function among current functions but does 
not mean that remote function registry does not hold even better matching 
function. To fix this issue we would first try to find function using 
`ExactFunctionResolver.getBestMatch()` and if exactly matching function is not 
found, we'll check if remote and local function registries are in sync and then 
use `DefaultFunctionResolver.getBestMatch()` to find the best matching 
function. But if exactly matching function is found, we'll return it right away 
without any registries sync checks.
    
    Changes:
    1. Add `consists` method to PersistentStore interface which can return true 
if key exists in store, false otherwise. This method is needed to return only 
remote function registry version without its content (unlike method `get`). 
We'll pull remote function registry content only if versions are different.
    2. Added check if remote and local function registries are in sync on any 
failure during planning stage
    and on exact matching function not found during execution stage.
    3. Added additional debug messages for `CreateFunctionHandler` and 
`DropFunctionHandler`.
    4. Updated unit tests to reflect new changes.



> Issues when overloading Drill native functions with dynamic UDFs
> ----------------------------------------------------------------
>
>                 Key: DRILL-4963
>                 URL: https://issues.apache.org/jira/browse/DRILL-4963
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Functions - Drill
>    Affects Versions: 1.9.0
>            Reporter: Roman
>            Assignee: Arina Ielchiieva
>             Fix For: Future
>
>         Attachments: subquery_udf-1.0.jar, subquery_udf-1.0-sources.jar, 
> test_overloading-1.0.jar, test_overloading-1.0-sources.jar
>
>
> I created jar file which overloads 3 DRILL native functions 
> (LOG(VARCHAR-REQUIRED), CURRENT_DATE(VARCHAR-REQUIRED) and 
> ABS(VARCHAR-REQUIRED,VARCHAR-REQUIRED)) and registered it as dynamic UDF.
> If I try to use my functions I will get errors:
> {code:xml}
> SELECT CURRENT_DATE('test') FROM (VALUES(1));
> {code}
> Error: FUNCTION ERROR: CURRENT_DATE does not support operand types (CHAR)
> SQL Query null
> {code:xml}
> SELECT ABS('test','test') FROM (VALUES(1));
> {code}
> Error: FUNCTION ERROR: ABS does not support operand types (CHAR,CHAR)
> SQL Query null
> {code:xml}
> SELECT LOG('test') FROM (VALUES(1));
> {code}
> Error: SYSTEM ERROR: DrillRuntimeException: Failure while materializing 
> expression in constant expression evaluator LOG('test').  Errors: 
> Error in expression at index -1.  Error: Missing function implementation: 
> castTINYINT(VARCHAR-REQUIRED).  Full expression: UNKNOWN EXPRESSION.
> But if I rerun all this queries after "DrillRuntimeException", they will run 
> correctly. It seems that Drill have not updated the function signature before 
> that error. Also if I add jar as usual UDF (copy jar to 
> /drill_home/jars/3rdparty and restart drillbits), all queries will run 
> correctly without errors.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to