aboutsummaryrefslogtreecommitdiffstats
path: root/python/sandcrawler
Commit message (Collapse)AuthorAgeFilesLines
* handle grobid2json errors in calling code insteadBryan Newbold2020-01-021-1/+7
|
* db: move duplicate row filtering into DB insert helpersBryan Newbold2020-01-022-15/+26
|
* remove unused filter in grobid workerBryan Newbold2020-01-021-1/+0
|
* fix dict typoBryan Newbold2020-01-021-1/+1
|
* improvements to grobid persist workerBryan Newbold2020-01-021-13/+16
|
* set mimetype when PUT to minioBryan Newbold2020-01-021-0/+4
|
* fix DB import countingBryan Newbold2020-01-021-4/+5
|
* fix small errors found by pylintBryan Newbold2020-01-022-1/+2
|
* fix sandcrawler persist workersBryan Newbold2020-01-021-0/+1
|
* filter ingest results to not have key conflicts within batchBryan Newbold2020-01-021-1/+16
| | | | | This handles a corner case with ON CONFLICT ... DO UPDATE where you can't do multiple such updates in the same batch transaction.
* db: fancy insert/update separation using postgres xmaxBryan Newbold2020-01-022-24/+45
|
* add PersistGrobidDiskWorkerBryan Newbold2020-01-021-0/+33
| | | | To help with making dumps directly from Kafka (eg, for partner delivery)
* flush out minio helper, add to grobid persistBryan Newbold2020-01-022-22/+71
|
* implement counts properly for persist workersBryan Newbold2020-01-021-15/+19
|
* improve DB helpersBryan Newbold2020-01-021-26/+81
| | | | | - return insert/update row counts - implement ON CONFLICT ... DO UPDATE on some tables
* be more parsimonious with GROBID metadataBryan Newbold2020-01-021-2/+4
| | | | | Because these are getting persisted in database (as well as kafka), don't write out empty keys.
* start work on DB connector and minio clientBryan Newbold2020-01-022-0/+200
|
* have JsonLinePusher continue on JSON decode errors (but count)Bryan Newbold2020-01-021-1/+5
|
* start work on persist workers and toolBryan Newbold2020-01-021-0/+223
|
* refactor: use print(..., file=sys.stderr)Bryan Newbold2019-12-183-25/+27
| | | | Should use logging soon, but this seems more idiomatic in the meanwhile.
* fixes for large GROBID result skipBryan Newbold2019-12-021-2/+2
|
* count empty blobs as 'failed' instead of crashingBryan Newbold2019-12-011-1/+2
| | | | Might be better to record an artificial kafka response instead?
* cleanup unused importBryan Newbold2019-12-011-1/+0
|
* filter out very large GROBID XML bodiesBryan Newbold2019-12-011-0/+6
| | | | | | | | | | This is to prevent Kafka MSG_SIZE_TOO_LARGE publish errors. We should probably bump this in the future. Open problems: hand-coding this size number isn't good, need to update in two places. Shouldn't filter out for non-Kafka sinks. Might still exist a corner-case where JSON encoded XML is larger than XML character string, due to encoding (eg, for unicode characters).
* CI: make some jobs manualBryan Newbold2019-11-151-0/+2
| | | | | Scalding test is broken :( But we aren't even using that code much these days.
* handle wayback fetch redirect loop in ingest codeBryan Newbold2019-11-141-2/+5
|
* bump kafka max poll interval for consumersBryan Newbold2019-11-141-2/+2
| | | | | The ingest worker keeps timing out at just over 5 minutes, so bump it just a bit.
* handle WaybackError during ingestBryan Newbold2019-11-141-0/+4
|
* handle SPNv1 redirect loopBryan Newbold2019-11-141-0/+2
|
* handle SPNv2 polling timeoutBryan Newbold2019-11-141-6/+10
|
* update ingest-file batch size to 1Bryan Newbold2019-11-141-3/+3
| | | | | | | | Was defaulting to 100, which I think was resulting in lots of consumer group timeouts, resulting in UNKNOWN_MEMBER_ID errors. Will probably switch back to batches of 10 or so, but multi-processing or some other concurrent dispatch/processing.
* start of hrmars.com ingest supportBryan Newbold2019-11-142-2/+7
|
* treat failure to get terminal capture as a SavePageNowErrorBryan Newbold2019-11-131-1/+1
|
* citation_pdf_url with host-relative URLsBryan Newbold2019-11-131-1/+3
|
* status_forcelist is on session, not requestBryan Newbold2019-11-131-2/+2
|
* handle SPNv1 remote server HTTP status codes betterBryan Newbold2019-11-131-8/+15
|
* handle requests (http) redirect loop from waybackBryan Newbold2019-11-131-1/+4
|
* handle wayback client return status correctlyBryan Newbold2019-11-131-2/+2
|
* allow way more errors in SPN pathBryan Newbold2019-11-131-2/+11
|
* clean up redirect-following CDX API pathBryan Newbold2019-11-131-8/+15
|
* fix lint errorsBryan Newbold2019-11-131-1/+1
|
* improve ingest worker remote failure behaviorBryan Newbold2019-11-131-5/+12
|
* have SPN client differentiate between SPN and remote errorsBryan Newbold2019-11-132-3/+11
| | | | | | | | This is only a partial implementation. The requests client will still make way too many SPN requests trying to figure out if this is a real error or not (eg, if remote was a 502, we'll retry many times). We may just want to switch to SPNv2 for everything.
* rename FileIngestWorkerBryan Newbold2019-11-132-5/+10
|
* refactor consume_topic name out of make_kafka_consumer()Bryan Newbold2019-11-131-5/+5
| | | | Best to do this in wrapping code for full flexibility.
* more progress on file ingestBryan Newbold2019-11-133-16/+73
|
* much progress on file ingest pathBryan Newbold2019-10-225-15/+334
|
* we do actually want consolidateHeader=2, not 1Bryan Newbold2019-10-041-3/+3
|
* grobid: consolidateHeaders typoBryan Newbold2019-10-041-1/+1
|
* disable citation consolidation by defaultBryan Newbold2019-10-041-1/+1
| | | | | | | with this consolidation enabled, the glutton_fatcat elasticsearch server was totally pegged over 90% CPU with only 10 PDF worker threads; the glutton load seemed to be the bottleneck even for this low degree of parallelism. Disabled for now, will debug with GROBID/glutton folks.
generated by cgit v1.2.3 (git 2.39.1) at 2024-05-22 22:18:43 +0000