scorch sorts fields by name to assign fieldID's

This is a stepping stone to allow easier future comparisons of field
maps and potential merge optimizations.

In bleve-blast tests on a 2015 macbook (50K wikipedia docs, 8
indexers, batch size 100, ssd), this does not seem to have a distinct
effect on indexing throughput.

index/scorch/segment/mem/build.go

@@ -95,6 +95,21 @@ func (s *Segment) initializeDict(results []*index.AnalysisResult) {
 	var numTokenFrequencies int
 	var totLocs int
 
+	// initial scan for all fieldID's to sort them
+	for _, result := range results {
+		for _, field := range result.Document.CompositeFields {
+			s.getOrDefineField(field.Name())
+		}
+		for _, field := range result.Document.Fields {
+			s.getOrDefineField(field.Name())
+		}
+	}
+	sort.Strings(s.FieldsInv[1:]) // keep _id as first field
+	s.FieldsMap = make(map[string]uint16, len(s.FieldsInv))
+	for fieldID, fieldName := range s.FieldsInv {
+		s.FieldsMap[fieldName] = uint16(fieldID + 1)
+	}
+
 	processField := func(fieldID uint16, tfs analysis.TokenFrequencies) {
 		for term, tf := range tfs {
 			pidPlus1, exists := s.Dicts[fieldID][term]

index/scorch/segment/zap/merge.go

@@ -21,6 +21,7 @@ import (
 	"fmt"
 	"math"
 	"os"
+	"sort"
 
 	"github.com/RoaringBitmap/roaring"
 	"github.com/Smerity/govarint"
@@ -545,5 +546,8 @@ func mergeFields(segments []*SegmentBase) []string {
 			rv = append(rv, k)
 		}
 	}
+
+	sort.Strings(rv[1:]) // leave _id as first
+
 	return rv
 }

index/scorch/segment/zap/segment_test.go

@@ -18,6 +18,7 @@ import (
 	"math"
 	"os"
 	"reflect"
+	"sort"
 	"testing"
 
 	"github.com/blevesearch/bleve/index"
@@ -574,6 +575,7 @@ func TestSegmentVisitableDocValueFieldsList(t *testing.T) {
 			t.Fatalf("segment VisitableDocValueFields err: %v", err)
 		}
 
+		sort.Strings(expectedFields[1:]) // keep _id as first field
 		if !reflect.DeepEqual(fields, expectedFields) {
 			t.Errorf("expected field terms: %#v, got: %#v", expectedFields, fields)
 		}