[
https://issues.apache.org/jira/browse/TIKA-5067?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Subbu updated TIKA-5067:
------------------------
Description:
Tika has no JSON magic or detector. The <html magic matches anywhere in the
first 8 KB of text, so a valid JSON document with an HTML tag inside a string
value is reported as text/html. This happens even when the file is named *.json
*Steps to reproduce:*
1) Check attached json
2) Run detect with the attached json.
*Actual:* text/html
*Expected:* application/json
*Fix options*
I see two ways to fix this: (A) a content-only JSON magic rule in
tika-mimetypes.xml, which is small but can't see the extension and would also
turn .txt, .csv and .js files that start like JSON into application/json, or
(B) a strict JSON check in MimeTypes.detect that overrides only a weak
text/html or text/plain verdict and respects the name and Content-Type hints.
Which approach do you prefer, or would you suggest another?
was:
Tika has no JSON magic or detector. The <html magic matches anywhere in the
first 8 KB of text, so a valid JSON document with an HTML tag inside a string
value is reported as text/html. This happens even when the file is named *.json
*Steps to reproduce:*
1) Check attached json
2) Run detect with the attached json.
*Actual:* text/html
*Expected:*
application/json
*Fix options*
I see two ways to fix this: (A) a content-only JSON magic rule in
tika-mimetypes.xml, which is small but can't see the extension and would also
turn .txt, .csv and .js files that start like JSON into application/json, or
(B) a strict JSON check in MimeTypes.detect that overrides only a weak
text/html or text/plain verdict and respects the name and Content-Type hints.
Which approach do you prefer, or would you suggest another?
> JSON with an HTML tag inside a string value is detected as text/html
> ---------------------------------------------------------------------
>
> Key: TIKA-5067
> URL: https://issues.apache.org/jira/browse/TIKA-5067
> Project: Tika
> Issue Type: Bug
> Components: detector
> Reporter: Subbu
> Priority: Major
> Attachments: findings.json
>
>
> Tika has no JSON magic or detector. The <html magic matches anywhere in the
> first 8 KB of text, so a valid JSON document with an HTML tag inside a string
> value is reported as text/html. This happens even when the file is named
> *.json
> *Steps to reproduce:*
> 1) Check attached json
> 2) Run detect with the attached json.
> *Actual:* text/html
>
> *Expected:* application/json
> *Fix options*
> I see two ways to fix this: (A) a content-only JSON magic rule in
> tika-mimetypes.xml, which is small but can't see the extension and would also
> turn .txt, .csv and .js files that start like JSON into application/json, or
> (B) a strict JSON check in MimeTypes.detect that overrides only a weak
> text/html or text/plain verdict and respects the name and Content-Type hints.
>
>
> Which approach do you prefer, or would you suggest another?
>
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)