aboutsummaryrefslogtreecommitdiffstats
path: root/skate/cmd/skate-reduce/main.go
blob: df72ef425d5417fe8d897137da8cff2636503c31 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
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)
		}
	}
}