[
https://issues.apache.org/jira/browse/TIKA-3510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17392352#comment-17392352
]
Thomas Mortagne commented on TIKA-3510:
---------------------------------------
bq. The goal was to make it possible for users to drop the sqlite3 jar and/or
the scientific-module jar on their path with the tika-parsers-standard-package
or tika-app/tika-server and have everything just work.
Yes, but it also makes it pretty much unusable in a project built with Maven
which uses Tika as library among other things which have common dependencies
but in different versions.
Even for the "drop the jar" use case you described, it can create quite a mess
if the environment where you drop it have conflicting dependencies (but I
supposed what you have in mind is only a very specific Tika server environment)
but you could improve that by shading those dependencies (rewrite the packages
in them) instead of just including them as is.
bq. One thing we could do is what we do in tika-parsers-standard and have a
module and package project for the scientific-module and the sqlite3 module.
We'd include the dependencies in the jar in *_package but not in *_module.
It's not super clear to me what tika-parsers-standard exactly is but in my
experience this kind of need is generally covered by providing 2 different
artifacts:
* tika-parser-scientific-module with regular Maven dependencies (so not
included)
* tika-parser-scientific-module-standalone (or something like this) with
everything in it
> tika-parser-scientific-module seems to embbed many dependencies
> ---------------------------------------------------------------
>
> Key: TIKA-3510
> URL: https://issues.apache.org/jira/browse/TIKA-3510
> Project: Tika
> Issue Type: Bug
> Components: parser
> Affects Versions: 2.0.0
> Reporter: Thomas Mortagne
> Priority: Major
>
> tika-parser-scientific-module 2.0.0 contains many files from other artifacts:
> * joda-time
> * slf4j
> * commons-io
> * ...
> Is that really expected ?
> tika-parser-sqlite3-module seems to be affected too
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
