diff options
author | Martin Czygan <martin.czygan@gmail.com> | 2021-05-06 19:47:35 +0200 |
---|---|---|
committer | Martin Czygan <martin.czygan@gmail.com> | 2021-05-06 19:47:35 +0200 |
commit | b46c9d351aa0a2a5c3618a1420259d4605a9654e (patch) | |
tree | 376f55f59baf657397210b2a2a3724b9ba61f6bb /skate/cmd/skate-reduce | |
parent | fb627564751d170c55b45785da569268b6290be9 (diff) | |
download | refcat-b46c9d351aa0a2a5c3618a1420259d4605a9654e.tar.gz refcat-b46c9d351aa0a2a5c3618a1420259d4605a9654e.zip |
move skate-verify to skate-reduce
Diffstat (limited to 'skate/cmd/skate-reduce')
-rw-r--r-- | skate/cmd/skate-reduce/main.go | 144 |
1 files changed, 144 insertions, 0 deletions
diff --git a/skate/cmd/skate-reduce/main.go b/skate/cmd/skate-reduce/main.go new file mode 100644 index 0000000..df72ef4 --- /dev/null +++ b/skate/cmd/skate-reduce/main.go @@ -0,0 +1,144 @@ +// skate-reduce takes prepared inputs (e.g. from skate-map or skate-cluster) +// and applies various verification and conversion functions. The output will +// often be the biblioref schema. +// +// Support various modes. +// +// * exact: takes (key, doc) TSV files (one for releases, one for refs) and +// will emit biblioref docs relating one element from releases with all +// elements from ref; this is for "doi", "pmid" and other id matches, where no +// further checks are necessary. The match reason, e.g. "doi" needs to be +// supplied. +// +// $ skate-reduce -m exact -r doi -F a.tsv -L b.tsv +// +// * verify: takes (key, doc) TSV files (one for release, one for refs), runs +// verification within a group and will emit biblioref. +// +// $ skate-reduce -m verify -F a.tsv -L b.tsv +// +// * ref: takes a single file with clusters containing releases and refs and +// will emit verification results. +// +// $ skate-reduce -m ref < a.ndj +// +// * bref: same as ref, but generate a biblioref file as output +// +// $ skate-reduce -m bref < a.ndj +// +// * wiki: zippy mode for releases and wikipedia inputs. +// +// $ skate-reduce -m wiki -L a.ndj -W b.ndj +// +package main + +import ( + "bufio" + "flag" + "log" + "os" + "runtime" + "runtime/pprof" + + "git.archive.org/martin/cgraph/skate" + "git.archive.org/martin/cgraph/skate/parallel" + "git.archive.org/martin/cgraph/skate/xio" +) + +var ( + numWorkers = flag.Int("w", runtime.NumCPU(), "number of workers") + batchSize = flag.Int("b", 10000, "batch size") + // Each mode may work on one or two files, and may need extra args. + mode = flag.String("m", "ref", "mode, e.g. exact, verify, ref, bref, wiki") + + cpuProfile = flag.String("cpuprofile", "", "write cpu profile to file") + memProfile = flag.String("memprofile", "", "write heap profile to file (go tool pprof -png --alloc_objects program mem.pprof > mem.png)") + + // Possible inputs, we could switch to a subcommand cli parser. + refs = flag.String("F", "", "path to refs input") + releases = flag.String("L", "", "path to release input") + wiki = flag.String("W", "", "path to wiki input") + + // Extra args. + reason = flag.String("r", "", "reason for match: doi, pmid, pmcid, arxiv, unknown") + reasonMap = map[string]skate.MatchResult{ + "doi": skate.MatchResult{Status: skate.StatusExact, Reason: skate.ReasonDOI}, + "pmid": skate.MatchResult{Status: skate.StatusExact, Reason: skate.ReasonPMID}, + "pmcid": skate.MatchResult{Status: skate.StatusExact, Reason: skate.ReasonPMCID}, + "arxiv": skate.MatchResult{Status: skate.StatusExact, Reason: skate.ReasonArxiv}, + "unknown": skate.MatchResult{Status: skate.StatusUnknown, Reason: skate.ReasonUnknown}, + } +) + +func main() { + flag.Parse() + if *cpuProfile != "" { + file, err := os.Create(*cpuProfile) + if err != nil { + log.Fatal(err) + } + pprof.StartCPUProfile(file) + defer pprof.StopCPUProfile() + } + + bw := bufio.NewWriter(os.Stdout) + defer bw.Flush() + + switch *mode { + case "exact": + l, f, err := xio.OpenTwo(*releases, *refs) + if err != nil { + log.Fatal(err) + } + r, ok := reasonMap[*reason] + if !ok { + log.Fatalf("unknown reason: %v", *reason) + } + if err := skate.ZippyExact(l, f, r, bw); err != nil { + log.Fatal(err) + } + case "verify": + l, f, err := xio.OpenTwo(*releases, *refs) + if err != nil { + log.Fatal(err) + } + if err := skate.ZippyVerifyRefs(l, f, bw); err != nil { + log.Fatal(err) + } + case "ref": + pp := parallel.NewProcessor(os.Stdin, os.Stdout, skate.RefClusterVerify) + pp.NumWorkers = *numWorkers + pp.BatchSize = *batchSize + if err := pp.Run(); err != nil { + log.Fatal(err) + } + case "bref": + pp := parallel.NewProcessor(os.Stdin, os.Stdout, skate.RefClusterToBiblioRef) + pp.NumWorkers = *numWorkers + pp.BatchSize = *batchSize + if err := pp.Run(); err != nil { + log.Fatal(err) + } + case "wiki": + l, w, err := xio.OpenTwo(*releases, *wiki) + if err != nil { + log.Fatal(err) + } + if err := skate.ZippyExactWiki(l, w, reasonMap["doi"], bw); err != nil { + log.Fatal(err) + } + default: + log.Fatalf("invalid mode") + } + if *memProfile != "" { + f, err := os.Create(*memProfile) + if err != nil { + log.Fatal("could not create memory profile: ", err) + } + defer f.Close() + runtime.GC() + if err := pprof.WriteHeapProfile(f); err != nil { + log.Fatal(err) + } + } +} |