diff options
| author | Bryan Newbold <bnewbold@archive.org> | 2020-01-21 10:59:27 -0800 |
|---|---|---|
| committer | Bryan Newbold <bnewbold@archive.org> | 2020-01-21 10:59:27 -0800 |
| commit | a1b44161e206873be30c0640f5fab7a284023ba1 (patch) | |
| tree | 9c5f2aa24a6d000e803b19427eb8a66730cd72be | |
| parent | fb7717ae410f72ff33017c176f64dff556b86f5b (diff) | |
| download | sandcrawler-a1b44161e206873be30c0640f5fab7a284023ba1.tar.gz sandcrawler-a1b44161e206873be30c0640f5fab7a284023ba1.zip | |
persist: work around GROBID timeouts with no status_code
| -rw-r--r-- | python/sandcrawler/db.py | 2 | ||||
| -rw-r--r-- | python/sandcrawler/persist.py | 4 |
2 files changed, 3 insertions, 3 deletions
diff --git a/python/sandcrawler/db.py b/python/sandcrawler/db.py index 3ec325e..e1414ba 100644 --- a/python/sandcrawler/db.py +++ b/python/sandcrawler/db.py @@ -161,7 +161,7 @@ class SandcrawlerPostgresClient: r['metadata'] = json.dumps(r['metadata'], sort_keys=True) batch = [(d['key'], d.get('grobid_version') or None, - d['status_code'], + d.get('status_code') or None, d['status'], d.get('fatcat_release') or None, d.get('updated') or datetime.datetime.now(), diff --git a/python/sandcrawler/persist.py b/python/sandcrawler/persist.py index 2463afa..801f76d 100644 --- a/python/sandcrawler/persist.py +++ b/python/sandcrawler/persist.py @@ -223,7 +223,7 @@ class PersistGrobidWorker(SandcrawlerWorker): self.counts['total'] += len(batch) for r in batch: - if r['status_code'] != 200 or not r.get('tei_xml'): + if r.get('status_code') != 200 or not r.get('tei_xml'): self.counts['s3-skip-status'] += 1 if r.get('error_msg'): r['metadata'] = {'error_msg': r['error_msg'][:500]} @@ -291,7 +291,7 @@ class PersistGrobidDiskWorker(SandcrawlerWorker): def process(self, record): - if record['status_code'] != 200 or not record.get('tei_xml'): + if record.get('status_code') != 200 or not record.get('tei_xml'): return False assert(len(record['key'])) == 40 p = "{}/{}".format(self.output_dir, self._blob_path(record['key'])) |