[
https://issues.apache.org/jira/browse/CTAKES-388?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15035387#comment-15035387
]
Lewis John McGibbney commented on CTAKES-388:
---------------------------------------------
[~fsqcds]
bq. As I understand access to such resources from jars will cause a slowdown.
The only issue with the previous File-based implementation was that we obtain a
NPE when the resource does not exist on disk at a particular location. This is
very difficult to maintain and scale when you attempt to run anything within a
cluster-based environment. That was the initial justification for moving this
into utilizing InputStream for streaming in the resources from a jar or in the
case of the cluster-based setup an Uber jar.
bq. Have you tried using UIMA's datapath for this?
No I have not. I was not even aware of the datapath API, can you please shine
some light on it? If it is a more appealing was of doing this then I suppose we
can undertake some tests and determine.
> URI is not hierarchical when attempting to obtain LookupDesc_Db.xml within JAR
> ------------------------------------------------------------------------------
>
> Key: CTAKES-388
> URL: https://issues.apache.org/jira/browse/CTAKES-388
> Project: cTAKES
> Issue Type: Improvement
> Components: ctakes-dictionary-lookup
> Affects Versions: 3.2.2
> Environment: MacOSX 10.9.5, cTAKES 3.2.3-SNAPSHOT, Apache Spark
> 1.5.1, Apache Hadoop 2.4.0
> Reporter: Lewis John McGibbney
> Fix For: 3.2.3
>
> Attachments: CTAKES-388.patch
>
>
> This issue is extremely similar to CTAKES-385 where the is an attempt to load
> a resource (in this case LookupDesc_Db.xml) as a File instead of an
> InputStream.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
