[jira] [Commented] (OPENNLP-1320) Makes lemmatize of MorfologikLemmatizer thread-safe

["ASF GitHub Bot (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OPENNLP-1320?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17628499#comment-17628499 ] 

ASF GitHub Bot commented on OPENNLP-1320:
-----------------------------------------

rzo1 commented on code in PR #385:
URL: https://github.com/apache/opennlp/pull/385#discussion_r1013330656


##########
opennlp-morfologik-addon/src/main/java/opennlp/morfologik/lemmatizer/MorfologikLemmatizer.java:
##########
@@ -47,7 +47,7 @@ public MorfologikLemmatizer(Dictionary dictionary) throws IllegalArgumentExcepti
     dictLookup = new DictionaryLookup(dictionary);
   }
 
-  private List<String> lemmatize(String word, String postag) {
+  private synchronized List<String> lemmatize(String word, String postag) {

Review Comment:
   An alternative to `synchronized` would be to re-create and dispose the `DictionaryLookup`, which is " is cheap to create and dispose (so it makes no sense to cache)" according to https://github.com/morfologik/morfologik-stemming/issues/69, i.e. moving to 
   
   ```java
   List<WordData> dictMap = new DictionaryLookup(dictionary).lookup(word.toLowerCase());
   ```
   
   and just store the (thread-safe) dictionary in the `MorfologikLemmatizer` instead of the not thread-safe `DictionaryLookup`. 
   
   That would avoid the synchronization cost. Wdyt?





> Makes lemmatize of MorfologikLemmatizer thread-safe
> ---------------------------------------------------
>
>                 Key: OPENNLP-1320
>                 URL: https://issues.apache.org/jira/browse/OPENNLP-1320
>             Project: OpenNLP
>          Issue Type: Bug
>            Reporter: Lucas Avanço
>            Priority: Major
>
> The method lemmatize of MorfologikLemmatizer is not thread-safe.
> Concurrent invokes may rise exceptions and return unpredictable resutls.
> It seems that the whole method must be sync because the variable returned by the morfologik lib is shared between threads.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)