diff options
| author | Bryan Newbold <bnewbold@archive.org> | 2022-10-24 10:19:17 -0700 |
|---|---|---|
| committer | Bryan Newbold <bnewbold@archive.org> | 2022-10-24 10:19:17 -0700 |
| commit | 97aa00038fbce39097d5f78b8891bbb88b71af75 (patch) | |
| tree | 853d27c41da25be16df9b257c0667efa7ae0f0d6 | |
| parent | 1657d4783a29eaee4adb610e0134a10b0126e202 (diff) | |
| download | sandcrawler-97aa00038fbce39097d5f78b8891bbb88b71af75.tar.gz sandcrawler-97aa00038fbce39097d5f78b8891bbb88b71af75.zip | |
html: pubpub platform detection
| -rw-r--r-- | python/sandcrawler/ingest_html.py | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/python/sandcrawler/ingest_html.py b/python/sandcrawler/ingest_html.py index 1c2c3fd..fb42e71 100644 --- a/python/sandcrawler/ingest_html.py +++ b/python/sandcrawler/ingest_html.py @@ -260,6 +260,8 @@ def html_guess_platform( in doc.html ): return "ojs" + if '<a href="https://www.pubpub.org">Published with' in doc.html: + return "pubpub" if 'Powered by <a target="_blank" href="http://arphahub.com">' in doc.html: return "arpha" if "<meta property='og:image' content='http://cms.galenos.com.tr' />" in doc.html: |