From 2069332aa58bd7d5804639a1adec3cde0118a5b0 Mon Sep 17 00:00:00 2001 From: Bryan Newbold Date: Tue, 25 Feb 2020 12:53:41 -0800 Subject: ingest: narrow xhtml filter --- python/sandcrawler/ingest.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'python') diff --git a/python/sandcrawler/ingest.py b/python/sandcrawler/ingest.py index 6ec54f6..0d4e7c6 100644 --- a/python/sandcrawler/ingest.py +++ b/python/sandcrawler/ingest.py @@ -290,7 +290,7 @@ class IngestFileWorker(SandcrawlerWorker): return result file_meta = gen_file_metadata(resource.body) - if "html" in file_meta['mimetype'] or "xml" in file_meta['mimetype']: + if "html" in file_meta['mimetype'] or "xhtml" in file_meta['mimetype'] or "application/xml" in file_meta['mimetype']: # Got landing page or similar. Some XHTML detected as "application/xml" if resource.terminal_dt: result['terminal'] = { -- cgit v1.2.3