cmd/go: exclude vendored packages from ... matches

By overwhelming popular demand, exclude vendored packages from ... matches,
by making ... never match the "vendor" element above a vendored package.

go help packages now reads:

    An import path is a pattern if it includes one or more "..." wildcards,
    each of which can match any string, including the empty string and
    strings containing slashes.  Such a pattern expands to all package
    directories found in the GOPATH trees with names matching the
    patterns.

    To make common patterns more convenient, there are two special cases.
    First, /... at the end of the pattern can match an empty string,
    so that net/... matches both net and packages in its subdirectories, like net/http.
    Second, any slash-separted pattern element containing a wildcard never
    participates in a match of the "vendor" element in the path of a vendored
    package, so that ./... does not match packages in subdirectories of
    ./vendor or ./mycode/vendor, but ./vendor/... and ./mycode/vendor/... do.
    Note, however, that a directory named vendor that itself contains code
    is not a vendored package: cmd/vendor would be a command named vendor,
    and the pattern cmd/... matches it.

Fixes #19090.

Change-Id: I985bf9571100da316c19fbfd19bb1e534a3c9e5f
Reviewed-on: https://go-review.googlesource.com/38745Reviewed-by: Alan Donovan <adonovan@google.com>

src/cmd/go/alldocs.go

@@ -316,7 +316,7 @@
 //
 // Usage:
 //
-// 	go env [var ...]
+// 	go env [-json] [var ...]
 //
 // Env prints Go environment information.
 //
@@ -325,6 +325,9 @@
 // names is given as arguments,  env prints the value of
 // each named variable on its own line.
 //
+// The -json flag prints the environment in JSON format
+// instead of as a shell script.
+//
 //
 // Start a bug report
 //
@@ -1361,8 +1364,19 @@
 // each of which can match any string, including the empty string and
 // strings containing slashes.  Such a pattern expands to all package
 // directories found in the GOPATH trees with names matching the
-// patterns.  As a special case, x/... matches x as well as x's subdirectories.
-// For example, net/... expands to net and packages in its subdirectories.
+// patterns.
+//
+// To make common patterns more convenient, there are two special cases.
+// First, /... at the end of the pattern can match an empty string,
+// so that net/... matches both net and packages in its subdirectories, like net/http.
+// Second, any slash-separated pattern element containing a wildcard never
+// participates in a match of the "vendor" element in the path of a vendored
+// package, so that ./... does not match packages in subdirectories of
+// ./vendor or ./mycode/vendor, but ./vendor/... and ./mycode/vendor/... do.
+// Note, however, that a directory named vendor that itself contains code
+// is not a vendored package: cmd/vendor would be a command named vendor,
+// and the pattern cmd/... matches it.
+// See golang.org/s/go15vendor for more about vendoring.
 //
 // An import path can also name a package to be downloaded from
 // a remote repository.  Run 'go help importpath' for details.

src/cmd/go/internal/help/helpdoc.go

@@ -71,8 +71,19 @@ An import path is a pattern if it includes one or more "..." wildcards,
 each of which can match any string, including the empty string and
 strings containing slashes.  Such a pattern expands to all package
 directories found in the GOPATH trees with names matching the
-patterns.  As a special case, x/... matches x as well as x's subdirectories.
-For example, net/... expands to net and packages in its subdirectories.
+patterns.
+
+To make common patterns more convenient, there are two special cases.
+First, /... at the end of the pattern can match an empty string,
+so that net/... matches both net and packages in its subdirectories, like net/http.
+Second, any slash-separated pattern element containing a wildcard never
+participates in a match of the "vendor" element in the path of a vendored
+package, so that ./... does not match packages in subdirectories of
+./vendor or ./mycode/vendor, but ./vendor/... and ./mycode/vendor/... do.
+Note, however, that a directory named vendor that itself contains code
+is not a vendored package: cmd/vendor would be a command named vendor,
+and the pattern cmd/... matches it.
+See golang.org/s/go15vendor for more about vendoring.
 
 An import path can also name a package to be downloaded from
 a remote repository.  Run 'go help importpath' for details.

src/cmd/go/internal/load/match_test.go

@@ -25,9 +25,43 @@ var matchPatternTests = `
 	match net net/http netchan
 	not not/http not/net/http
 	
+	# Special cases. Quoting docs:
+
+	# First, /... at the end of the pattern can match an empty string,
+	# so that net/... matches both net and packages in its subdirectories, like net/http.
 	pattern net/...
 	match net net/http
 	not not/http not/net/http netchan
+
+	# Second, any slash-separted pattern element containing a wildcard never
+	# participates in a match of the "vendor" element in the path of a vendored
+	# package, so that ./... does not match packages in subdirectories of
+	# ./vendor or ./mycode/vendor, but ./vendor/... and ./mycode/vendor/... do.
+	# Note, however, that a directory named vendor that itself contains code
+	# is not a vendored package: cmd/vendor would be a command named vendor,
+	# and the pattern cmd/... matches it.
+	pattern ./...
+	match ./vendor ./mycode/vendor
+	not ./vendor/foo ./mycode/vendor/foo
+	
+	pattern ./vendor/...
+	match ./vendor/foo ./vendor/foo/vendor
+	not ./vendor/foo/vendor/bar
+	
+	pattern mycode/vendor/...
+	match mycode/vendor mycode/vendor/foo mycode/vendor/foo/vendor
+	not mycode/vendor/foo/vendor/bar
+	
+	pattern x/vendor/y
+	match x/vendor/y
+	not x/vendor
+	
+	pattern x/vendor/y/...
+	match x/vendor/y x/vendor/y/z x/vendor/y/vendor x/vendor/y/z/vendor
+	not x/vendor/y/vendor/z
+	
+	pattern .../vendor/...
+	match x/vendor/y x/vendor/y/z x/vendor/y/vendor x/vendor/y/z/vendor
 `
 
 func TestMatchPattern(t *testing.T) {

src/cmd/go/internal/load/search.go

@@ -202,17 +202,69 @@ func treeCanMatchPattern(pattern string) func(name string) bool {
 // name matches pattern. Pattern is a limited glob
 // pattern in which '...' means 'any string' and there
 // is no other special syntax.
+// Unfortunately, there are two special cases. Quoting "go help packages":
+//
+// First, /... at the end of the pattern can match an empty string,
+// so that net/... matches both net and packages in its subdirectories, like net/http.
+// Second, any slash-separted pattern element containing a wildcard never
+// participates in a match of the "vendor" element in the path of a vendored
+// package, so that ./... does not match packages in subdirectories of
+// ./vendor or ./mycode/vendor, but ./vendor/... and ./mycode/vendor/... do.
+// Note, however, that a directory named vendor that itself contains code
+// is not a vendored package: cmd/vendor would be a command named vendor,
+// and the pattern cmd/... matches it.
 func matchPattern(pattern string) func(name string) bool {
-	re := regexp.QuoteMeta(pattern)
-	re = strings.Replace(re, `\.\.\.`, `.*`, -1)
-	// Special case: foo/... matches foo too.
-	if strings.HasSuffix(re, `/.*`) {
-		re = re[:len(re)-len(`/.*`)] + `(/.*)?`
+	// Convert pattern to regular expression.
+	// The strategy for the trailing /... is to nest it in an explicit ? expression.
+	// The strategy for the vendor exclusion is to change the unmatchable
+	// vendor strings to a disallowed code point (vendorChar) and to use
+	// "(anything but that codepoint)*" as the implementation of the ... wildcard.
+	// This is a bit complicated but the obvious alternative,
+	// namely a hand-written search like in most shell glob matchers,
+	// is too easy to make accidentally exponential.
+	// Using package regexp guarantees linear-time matching.
+
+	const vendorChar = "\x00"
+
+	if strings.Contains(pattern, vendorChar) {
+		return func(name string) bool { return false }
 	}
+
+	re := regexp.QuoteMeta(pattern)
+	re = replaceVendor(re, vendorChar)
+	switch {
+	case strings.HasSuffix(re, `/`+vendorChar+`/\.\.\.`):
+		re = strings.TrimSuffix(re, `/`+vendorChar+`/\.\.\.`) + `(/vendor|/` + vendorChar + `/\.\.\.)`
+	case re == vendorChar+`/\.\.\.`:
+		re = `(/vendor|/` + vendorChar + `/\.\.\.)`
+	case strings.HasSuffix(re, `/\.\.\.`):
+		re = strings.TrimSuffix(re, `/\.\.\.`) + `(/\.\.\.)?`
+	}
+	re = strings.Replace(re, `\.\.\.`, `[^`+vendorChar+`]*`, -1)
+
 	reg := regexp.MustCompile(`^` + re + `$`)
+
 	return func(name string) bool {
-		return reg.MatchString(name)
+		if strings.Contains(name, vendorChar) {
+			return false
+		}
+		return reg.MatchString(replaceVendor(name, vendorChar))
+	}
+}
+
+// replaceVendor returns the result of replacing
+// non-trailing vendor path elements in x with repl.
+func replaceVendor(x, repl string) string {
+	if !strings.Contains(x, "vendor") {
+		return x
+	}
+	elem := strings.Split(x, "/")
+	for i := 0; i < len(elem)-1; i++ {
+		if elem[i] == "vendor" {
+			elem[i] = repl
+		}
 	}
+	return strings.Join(elem, "/")
 }
 
 // ImportPaths returns the import paths to use for the given command line.

src/cmd/go/vendor_test.go

@@ -20,18 +20,18 @@ func TestVendorImports(t *testing.T) {
 	tg := testgo(t)
 	defer tg.cleanup()
 	tg.setenv("GOPATH", filepath.Join(tg.pwd(), "testdata"))
-	tg.run("list", "-f", "{{.ImportPath}} {{.Imports}}", "vend/...")
+	tg.run("list", "-f", "{{.ImportPath}} {{.Imports}}", "vend/...", "vend/vendor/...", "vend/x/vendor/...")
 	want := `
 		vend [vend/vendor/p r]
 		vend/dir1 []
 		vend/hello [fmt vend/vendor/strings]
 		vend/subdir [vend/vendor/p r]
+		vend/x [vend/x/vendor/p vend/vendor/q vend/x/vendor/r vend/dir1 vend/vendor/vend/dir1/dir2]
+		vend/x/invalid [vend/x/invalid/vendor/foo]
 		vend/vendor/p []
 		vend/vendor/q []
 		vend/vendor/strings []
 		vend/vendor/vend/dir1/dir2 []
-		vend/x [vend/x/vendor/p vend/vendor/q vend/x/vendor/r vend/dir1 vend/vendor/vend/dir1/dir2]
-		vend/x/invalid [vend/x/invalid/vendor/foo]
 		vend/x/vendor/p []
 		vend/x/vendor/p/p [notfound]
 		vend/x/vendor/r []