[
https://issues.apache.org/jira/browse/TIKA-5037?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18138894#comment-18138894
]
ASF GitHub Bot commented on TIKA-5037:
--------------------------------------
dschmidt commented on PR #3531:
URL: https://github.com/apache/tika/pull/3531#issuecomment-6663588052
Thanks, all addressed in the last commit.
1. Quadratic string building: an error output now keeps its first `ename`
and `evalue` and joins them once. A test with 100,000 repeated `ename` keys
runs in well under a second.
2. `supported-formats.adoc` is regenerated with the `JupyterParser` entry;
`testSupportedFormatsPartialIsCurrent` passes.
- Merge conflicts: resolved earlier today when merging main.
- Tests: the document and string limits are covered on `Notebook.read` (with
buffers large enough for Jackson to check them), and bad base64 is recorded and
skipped. That test turned up a real gap: `Base64.getMimeDecoder()` silently
skips characters that are not base64, so garbage decoded into a broken image.
The data is now decoded strictly once whitespace is removed; 18 real notebooks
still decode every image without an error.
- The long line in `formats.adoc` is rewrapped.
> Jupyter notebooks: parse .ipynb, emit the first image output as THUMBNAIL
> -------------------------------------------------------------------------
>
> Key: TIKA-5037
> URL: https://issues.apache.org/jira/browse/TIKA-5037
> Project: Tika
> Issue Type: New Feature
> Components: mime, parser
> Reporter: Dominik Schmidt
> Priority: Major
>
> There is no media type or parser for Jupyter notebooks today;
> shared-mime-info uses {{application/x-ipynb+json}}. A parser could extract
> the cell sources as text, and the cell outputs carry base64 {{image/png}}
> data, the first of which would make a useful THUMBNAIL.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
