bio-duplication-stats: use maxX and maxY computed by bio-mark-duplicates

Summary: This revision outputs max X coordinate and max Y coordinate computed by bio-mark-duplicates in a json file as tile width and height respectively. bio-duplication stats uses them for global coordinate computations.

Test Plan: Unit tests and tiny targeted test.

Reviewers: ayip, escott, kshashidhar

Reviewed By: ayip

Maniphest Tasks: T35065

Differential Revision: https://phabricator.grailbio.com/D44038

fbshipit-source-id: d2eaf3d

Author: akokate

Diff for: main.go

@@ -39,6 +39,7 @@ var (
 	format               = flag.String("format", "bam", "Output format. Value is either 'bam' or 'pam'.")
 	metricsFile          = flag.String("metrics", "", "Output metrics file")
 	highCovFile          = flag.String("high-cov-regions", "", "Output high coverage regions file")
+	tileSizeFile         = flag.String("tile-size", "", "Output width and height of tile to file")
 	scratchDir           = flag.String("scratch-dir", "/tmp", "Directory to put scratch files")
 	parallelism          = flag.Int("parallelism", runtime.NumCPU(), "Number of parallel computations to run during the markdup phase")
 	queueLength          = flag.Int("queue-length", runtime.NumCPU()*5, "Number shards to queue while waiting for flush")
@@ -79,6 +80,7 @@ func main() {
 		IndexFile:                *indexFile,
 		MetricsFile:              *metricsFile,
 		HighCoverageIntervalFile: *highCovFile,
+		TileSizeFile:             *tileSizeFile,
 		Format:                   *format,
 		CoverageMax:              *maxDepth,
 		ShardSize:                *shardSize,

Diff for: markduplicates/mark_duplicates.go

@@ -61,9 +61,9 @@ type OpticalDetector interface {
 	// influence optical detection.
 	GetRecordProcessor() bampair.RecordProcessor
 
-	// RecordProcessorsDone should be called after the RecordProcessors
-	// have seen all the input records.
-	RecordProcessorsDone()
+	// RecordProcessorsDone should return maximum X and Y co-ordinates and
+	// be called after the RecordProcessors have seen all the input records.
+	RecordProcessorsDone() (int, int)
 
 	// Detect identifies the optical duplicates in pairs and returns
 	// their names in a slice. readGroupLibrary maps readGroup to
@@ -80,6 +80,7 @@ type Opts struct {
 	IndexFile                string
 	MetricsFile              string
 	HighCoverageIntervalFile string
+	TileSizeFile             string
 	Format                   string
 	CoverageMax              int
 	ShardSize                int
@@ -246,7 +247,7 @@ func (m *MarkDuplicates) Mark(shards []bam.Shard) (*MetricsCollection, error) {
 	m.shardInfo = shardInfo
 	m.globalMetrics.maxAlignDist = m.globalMaxAlignDist
 	if m.Opts.OpticalDetector != nil {
-		m.Opts.OpticalDetector.RecordProcessorsDone()
+		m.globalMetrics.maxX, m.globalMetrics.maxY = m.Opts.OpticalDetector.RecordProcessorsDone()
 	}
 
 	// Determine high coverage intervals if desired.
@@ -874,6 +875,11 @@ func SetupAndMark(ctx context.Context, provider bamprovider.Provider, opts *Opts
 			return err
 		}
 	}
+	if opts.TileSizeFile != "" {
+		if err := writeTileSize(ctx, opts, globalMetrics); err != nil {
+			return err
+		}
+	}
 	if opts.OpticalHistogram != "" {
 		if err := writeOpticalHistogram(ctx, opts, globalMetrics); err != nil {
 			return err

Diff for: markduplicates/metrics.go

@@ -15,6 +15,7 @@ package markduplicates
 
 import (
 	"context"
+	"encoding/json"
 	"fmt"
 	"os"
 	"sort"
@@ -94,6 +95,8 @@ func (m *Metrics) Add(other *Metrics) {
 type MetricsCollection struct {
 	// Global metrics
 	maxAlignDist int
+	maxX         int
+	maxY         int
 
 	// OpticalDistance stores the number of duplicate read pairs that
 	// have the given Euclidean distance.
@@ -253,6 +256,24 @@ func writeHighCoverageIntervals(ctx context.Context, opts *Opts, header *sam.Hea
 	return nil
 }
 
+func writeTileSize(ctx context.Context, opts *Opts, globalMetrics *MetricsCollection) (err error) {
+	var f *os.File
+	f, err = os.Create(opts.TileSizeFile)
+	if err != nil {
+		return errors.E(err, "Couldn't create tile size file:", opts.TileSizeFile)
+	}
+	defer func() {
+		if err2 := f.Close(); err == nil && err2 != nil {
+			err = err2
+		}
+	}()
+	enc := json.NewEncoder(f)
+	return enc.Encode(map[string]int{
+		"tileWidth":  globalMetrics.maxX,
+		"tileHeight": globalMetrics.maxY,
+	})
+}
+
 func writeOpticalHistogram(ctx context.Context, opts *Opts, globalMetrics *MetricsCollection) (err error) {
 	var f *os.File
 	f, err = os.Create(opts.OpticalHistogram)

Diff for: markduplicates/optical_detector.go

@@ -98,7 +98,8 @@ func (t *TileOpticalDetector) GetRecordProcessor() bampair.RecordProcessor {
 }
 
 // RecordProcessorsDone implements OpticalDetector.
-func (t *TileOpticalDetector) RecordProcessorsDone() {
+func (t *TileOpticalDetector) RecordProcessorsDone() (int, int) {
+	return 0, 0
 }
 
 // Detect implements OpticalDetector.