[ https://issues.apache.org/jira/browse/TIKA-645?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13036214#comment-13036214 ]
Nick Burch commented on TIKA-645: --------------------------------- Thanks. With this change in, tika-app from svn trunk can parse the same 5mb excel file with "-Xms15m -Xmx15m", so the use of NPOIFS and access to the file has saved us a bit over the size of the office file in memory, which is good news! > Parsers can't get at an underlying TikaInputStream to get the file if they > wanted one > ------------------------------------------------------------------------------------- > > Key: TIKA-645 > URL: https://issues.apache.org/jira/browse/TIKA-645 > Project: Tika > Issue Type: Bug > Components: parser > Affects Versions: 0.9 > Reporter: Nick Burch > Assignee: Jukka Zitting > Fix For: 1.0 > > > Spotted this with the office parser, but it should be general. The user > creates a TikaInputStream, and passes that off to the parser framework. The > Parser that is called may wish to spot that the input is a File backed > TikaInputStream, and take a shortcut to use the file instead of the > InputStream. > However, what the parser gets is a TaggedInputStream wrapping a > CountingInputStream wrapping the original TikaInputStream. As such, it can't > get at the file. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira