[tools][symbolize] Make text parser more robust
Currently text will never parse past "{{{" or "\033[".
This causes issues if you have partial versions of either
these embeded in otherwise meaningful lines. So we'd like
to not give up on the whole line just because one part
looks a bit funny.
This change makes text consume everything upto a regular
expression which matches the parts of text that can actully
be parsed.
Change-Id: I5ed22d6d8cf5cc53fe5dcc5687ec87e13e9144e9
diff --git a/symbolize/filter_test.go b/symbolize/filter_test.go
index ea40b66..6b524da 100644
--- a/symbolize/filter_test.go
+++ b/symbolize/filter_test.go
@@ -111,9 +111,9 @@
func TestMalformed(t *testing.T) {
// Parse a bad line
line := ParseLine("\033[1m Error at {{{pc:0x123879c0")
-
- if line != nil {
- t.Error("expected", nil, "got", line)
+ // Malformed lines should still parse
+ if line == nil {
+ t.Error("expected", "not nil", "got", line)
}
}
diff --git a/symbolize/parser.go b/symbolize/parser.go
index d8988a8..9e460ae 100644
--- a/symbolize/parser.go
+++ b/symbolize/parser.go
@@ -10,11 +10,9 @@
"fmt"
"io"
"regexp"
- "strings"
)
const (
- elemPrefix string = "{{{"
elemSuffix string = "}}}"
colorPrefix string = "\033["
modulePrefix string = "{{{module:"
@@ -23,21 +21,18 @@
btPrefix string = "{{{bt:"
)
-func findIndex(s ParserState, sub string) int {
- idx := strings.Index(string(s), sub)
- if idx == -1 {
- return len(s)
- }
- return idx
-}
+var (
+ endTextRegex = regexp.MustCompile("({{{.*}}})|(\033\\[[0-9]+m)")
+ beginLogLineRegex = regexp.MustCompile(`\[[0-9]+\.[0-9]+\] [0-9]+\.[0-9]+>`)
+)
func ParseText(b *ParserState) interface{} {
- idx := findIndex(*b, elemPrefix)
- idx2 := findIndex(*b, colorPrefix)
- if idx2 < idx {
- idx = idx2
+ var idx int
+ if loc := endTextRegex.FindStringIndex(string(*b)); loc != nil {
+ idx = loc[0]
+ } else {
+ idx = len(*b)
}
-
if idx == 0 {
return nil
}
@@ -189,7 +184,6 @@
out := make(chan InputLine)
// This is not used for demuxing. It is a human readable line number.
var lineno uint64 = 1
- re := regexp.MustCompile(`\[[0-9]+\.[0-9]+\] [0-9]+\.[0-9]+>`)
go func() {
defer close(out)
scanner := bufio.NewScanner(reader)
@@ -202,7 +196,7 @@
text := ParserState(scanner.Text())
b := &text
// Get the dummyText and needed text.
- locs := re.FindStringIndex(string(text))
+ locs := beginLogLineRegex.FindStringIndex(string(text))
if locs == nil {
// This means the whole thing is dummy text.
var line InputLine
diff --git a/symbolize/parserstate.go b/symbolize/parserstate.go
index bb8a57f..2dfd2e4 100644
--- a/symbolize/parserstate.go
+++ b/symbolize/parserstate.go
@@ -29,17 +29,6 @@
return false
}
-// This is like before but does not consume the expected delimiter
-func (b *ParserState) onlyBefore(what string) (string, error) {
- idx := strings.Index(string(*b), what)
- if idx == -1 {
- return "", fmt.Errorf("expected '%s'", what)
- }
- str := (*b)[:idx]
- *b = (*b)[idx:]
- return string(str), nil
-}
-
func (b *ParserState) before(what string) (string, error) {
idx := strings.Index(string(*b), what)
if idx == -1 {
diff --git a/symbolize/parserstate_test.go b/symbolize/parserstate_test.go
index 638ca3f..f1a8728 100644
--- a/symbolize/parserstate_test.go
+++ b/symbolize/parserstate_test.go
@@ -38,30 +38,6 @@
}
}
-func TestOnlyBefore(t *testing.T) {
- buf := ParserState("this is a test")
- v1, err := buf.onlyBefore(" ")
- if err != nil {
- t.Error(err)
- }
- if v1 != "this" {
- t.Error("expected", "this", "got", v1)
- }
- if string(buf) != " is a test" {
- t.Error("expected", " is a test", "got", string(buf))
- }
- // The following is just to consume the trailing space
- buf.before(" ")
-
- _, err = buf.onlyBefore("#")
- if err == nil {
- t.Error("expected an error but got none")
- }
- if string(buf) != "is a test" {
- t.Error("input consumed when it should not have been")
- }
-}
-
func TestBefore(t *testing.T) {
buf := ParserState("this is a test")
v1, err := buf.before(" ")
