Support platform file paths through escape

Signed-off-by: John Howard <jhoward@microsoft.com>
Upstream-commit: e8e3dd32c5bad727010ec787f484b98942977531
Component: engine
This commit is contained in:
John Howard
2016-04-22 15:04:46 -07:00
parent 421a9caea1
commit 886f3ce26b
13 changed files with 279 additions and 32 deletions

View File

@ -94,12 +94,12 @@ func parseWords(rest string) []string {
blankOK = true
phase = inQuote
}
if ch == '\\' {
if ch == tokenEscape {
if pos+1 == len(rest) {
continue // just skip \ at end
continue // just skip an escape token at end of line
}
// If we're not quoted and we see a \, then always just
// add \ plus the char to the word, even if the char
// If we're not quoted and we see an escape token, then always just
// add the escape token plus the char to the word, even if the char
// is a quote.
word += string(ch)
pos++
@ -112,11 +112,11 @@ func parseWords(rest string) []string {
if ch == quote {
phase = inWord
}
// \ is special except for ' quotes - can't escape anything for '
if ch == '\\' && quote != '\'' {
// The escape token is special except for ' quotes - can't escape anything for '
if ch == tokenEscape && quote != '\'' {
if pos+1 == len(rest) {
phase = inWord
continue // just skip \ at end
continue // just skip the escape token at end
}
pos++
nextCh := rune(rest[pos])

View File

@ -3,6 +3,7 @@ package parser
import (
"bufio"
"fmt"
"io"
"regexp"
"strings"
@ -37,10 +38,26 @@ type Node struct {
var (
dispatch map[string]func(string) (*Node, map[string]bool, error)
tokenWhitespace = regexp.MustCompile(`[\t\v\f\r ]+`)
tokenLineContinuation = regexp.MustCompile(`\\[ \t]*$`)
tokenLineContinuation *regexp.Regexp
tokenEscape rune
tokenEscapeCommand = regexp.MustCompile(`^#[ \t]*escape[ \t]*=[ \t]*(?P<escapechar>.).*$`)
tokenComment = regexp.MustCompile(`^#.*$`)
lookingForDirectives bool
directiveEscapeSeen bool
)
const defaultTokenEscape = "\\"
// setTokenEscape sets the default token for escaping characters in a Dockerfile.
func setTokenEscape(s string) error {
if s != "`" && s != "\\" {
return fmt.Errorf("invalid ESCAPE '%s'. Must be ` or \\", s)
}
tokenEscape = rune(s[0])
tokenLineContinuation = regexp.MustCompile(`\` + s + `[ \t]*$`)
return nil
}
func init() {
// Dispatch Table. see line_parsers.go for the parse functions.
// The command is parsed and mapped to the line parser. The line parser
@ -70,6 +87,29 @@ func init() {
// ParseLine parse a line and return the remainder.
func ParseLine(line string) (string, *Node, error) {
// Handle the parser directive '# escape=<char>. Parser directives must preceed
// any builder instruction or other comments, and cannot be repeated.
if lookingForDirectives {
tecMatch := tokenEscapeCommand.FindStringSubmatch(strings.ToLower(line))
if len(tecMatch) > 0 {
if directiveEscapeSeen == true {
return "", nil, fmt.Errorf("only one escape parser directive can be used")
}
for i, n := range tokenEscapeCommand.SubexpNames() {
if n == "escapechar" {
if err := setTokenEscape(tecMatch[i]); err != nil {
return "", nil, err
}
directiveEscapeSeen = true
return "", nil, nil
}
}
}
}
lookingForDirectives = false
if line = stripComments(line); line == "" {
return "", nil, nil
}
@ -103,6 +143,9 @@ func ParseLine(line string) (string, *Node, error) {
// Parse is the main parse routine.
// It handles an io.ReadWriteCloser and returns the root of the AST.
func Parse(rwc io.Reader) (*Node, error) {
directiveEscapeSeen = false
lookingForDirectives = true
setTokenEscape(defaultTokenEscape) // Assume the default token for escape
currentLine := 0
root := &Node{}
root.StartLine = -1

View File

@ -131,22 +131,22 @@ func TestLineInformation(t *testing.T) {
t.Fatalf("Error parsing dockerfile %s: %v", testFileLineInfo, err)
}
if ast.StartLine != 4 || ast.EndLine != 30 {
fmt.Fprintf(os.Stderr, "Wrong root line information: expected(%d-%d), actual(%d-%d)\n", 4, 30, ast.StartLine, ast.EndLine)
if ast.StartLine != 5 || ast.EndLine != 31 {
fmt.Fprintf(os.Stderr, "Wrong root line information: expected(%d-%d), actual(%d-%d)\n", 5, 31, ast.StartLine, ast.EndLine)
t.Fatalf("Root line information doesn't match result.")
}
if len(ast.Children) != 3 {
fmt.Fprintf(os.Stderr, "Wrong number of child: expected(%d), actual(%d)\n", 3, len(ast.Children))
t.Fatalf("Root line information doesn't match result.")
t.Fatalf("Root line information doesn't match result for %s", testFileLineInfo)
}
expected := [][]int{
{4, 4},
{10, 11},
{16, 30},
{5, 5},
{11, 12},
{17, 31},
}
for i, child := range ast.Children {
if child.StartLine != expected[i][0] || child.EndLine != expected[i][1] {
fmt.Fprintf(os.Stderr, "Wrong line information for child %d: expected(%d-%d), actual(%d-%d)\n",
t.Logf("Wrong line information for child %d: expected(%d-%d), actual(%d-%d)\n",
i, expected[i][0], expected[i][1], child.StartLine, child.EndLine)
t.Fatalf("Root line information doesn't match result.")
}

View File

@ -1,3 +1,4 @@
#escape=\
FROM brimstone/ubuntu:14.04
MAINTAINER brimstone@the.narro.ws

View File

@ -0,0 +1,9 @@
# Comment here. Should not be looking for the following parser directive.
# Hence the following line will be ignored, and the subsequent backslash
# continuation will be the default.
# escape = `
FROM image
MAINTAINER foo@bar.com
ENV GOPATH \
\go

View File

@ -0,0 +1,3 @@
(from "image")
(maintainer "foo@bar.com")
(env "GOPATH" "\\go")

View File

@ -0,0 +1,7 @@
# escape = ``
# There is no white space line after the directives. This still succeeds, but goes
# against best practices.
FROM image
MAINTAINER foo@bar.com
ENV GOPATH `
\go

View File

@ -0,0 +1,3 @@
(from "image")
(maintainer "foo@bar.com")
(env "GOPATH" "\\go")

View File

@ -0,0 +1,6 @@
#escape = `
FROM image
MAINTAINER foo@bar.com
ENV GOPATH `
\go

View File

@ -0,0 +1,3 @@
(from "image")
(maintainer "foo@bar.com")
(env "GOPATH" "\\go")