diff options
| author | Bryan Newbold <bnewbold@archive.org> | 2020-09-14 14:13:34 -0700 |
|---|---|---|
| committer | Bryan Newbold <bnewbold@archive.org> | 2020-09-14 14:13:34 -0700 |
| commit | ee6129ea884036b666de7cff4ad7891675a52b3c (patch) | |
| tree | f3f2d4970f2622b16425eab7ae0de2eacac30ef5 | |
| parent | 62252a6179953ccc79a6cb60c40a756fa0a034e1 (diff) | |
| download | sandcrawler-ee6129ea884036b666de7cff4ad7891675a52b3c.tar.gz sandcrawler-ee6129ea884036b666de7cff4ad7891675a52b3c.zip | |
ingest: treat text/xml as XHTML in pdf ingest
| -rw-r--r-- | python/sandcrawler/ingest.py | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/python/sandcrawler/ingest.py b/python/sandcrawler/ingest.py index c45437d..e8b3551 100644 --- a/python/sandcrawler/ingest.py +++ b/python/sandcrawler/ingest.py @@ -429,7 +429,7 @@ class IngestFileWorker(SandcrawlerWorker): ) file_meta = gen_file_metadata(resource.body) - if "html" in file_meta['mimetype'] or "xhtml" in file_meta['mimetype'] or "application/xml" in file_meta['mimetype']: + if "html" in file_meta['mimetype'] or "xhtml" in file_meta['mimetype'] or "application/xml" in file_meta['mimetype'] or "text/xml" in file_meta['mimetype']: # Got landing page or similar. Some XHTML detected as "application/xml" if resource.terminal_dt: result['terminal'] = { |