[ https://issues.apache.org/jira/browse/TIKA-2478?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16214491#comment-16214491 ]
Ken Krugler commented on TIKA-2478: ----------------------------------- I recently had to dig into extracting text from emails, and it isn't all that straightforward. E.g. you _do_ want to combine text from {{multipart/mixed}}, but not {{multipart/alternative}}, where you generally want to favor text over HTML. There's also {{multipart/related}} and {{multipart/signed}}. And each {{multipart/mixed}} piece has to be evaluated to extract (potential) text; typically where I've seen this is with an inlined image, so you get text/html, image/whatever, text/html. > MBOX import includes redundant copies of the text > ------------------------------------------------- > > Key: TIKA-2478 > URL: https://issues.apache.org/jira/browse/TIKA-2478 > Project: Tika > Issue Type: Bug > Affects Versions: 1.16 > Reporter: Robert Letzler > Priority: Minor > > MBOX messages often get parsed into four documents: > a. The mbox file - outer container "/" > b. The actual email-- "/embedded-1" > c. The utf-8 text content of the email "/embedded-1/embedded-2" > d. The utf-8 html content of the email "/embedded-1/embedded-3" > entries C and D are redundant and distracting. The MSG parser parses the > first non-null: email body and then it skips the rest. Please modify MBOX to > not have separate "attached" documents for the html body and the text body. > The attachment to https://issues.apache.org/jira/browse/TIKA-2471 is an > example of input sufficient to generate this behavior. > Thanks! -- This message was sent by Atlassian JIRA (v6.4.14#64029)