The import path of this package has changed in PR #30. /cc @sqs

However, the latest released version (v0.5.0) still has the old module path in its go.mod file:

https://github.com/sourcegraph/go-diff/blob/v0.5.0/go.mod#L1

As a result, trying to install the latest version of this package in module mode fails:

$ cd $(mktemp -d)

$ go mod init m
go: creating new go.mod: module m

$ go get github.com/sourcegraph/go-diff/diff
go: finding github.com/sourcegraph/go-diff/diff latest
go: github.com/sourcegraph/[email protected]: parsing go.mod: unexpected module path "sourcegraph.com/sourcegraph/go-diff"
go: error loading module requirements

$ echo $?
1

This can be fixed this by making a new release version (perhaps v0.5.1) that contains the new module path in the go.mod file.

This is a campaign run to fulfill the delivery plan https://github.com/sourcegraph/customer/issues/13. Over the coming days, I will update more repositories using go mod tidy.

It seems that the fractional seconds part is optional, so change the format to accept whole seconds too.

Here's a sample diff output by the bzr command which fails to parse currently.

I started adding a test, but it involves more time than I have to spend right now, as the tests rely on reflect.DeepEqual working on time.Time instances, which won't work in general due to monotonic time. I'd suggest using https://godoc.org/github.com/google/go-cmp/cmp#Diff instead.

=== modified file 'encode.go'
--- encode.go	2011-11-24 19:47:20 +0000
+++ encode.go	2011-11-28 13:24:31 +0000
@@ -250,8 +250,7 @@
 	e.emitScalar("null", "", "", C.YAML_PLAIN_SCALAR_STYLE)
 }

-func (e *encoder) emitScalar(value, anchor, tag string,
-	style C.yaml_scalar_style_t) {
+func (e *encoder) emitScalar(value, anchor, tag string, style C.yaml_scalar_style_t) {
 	var canchor, ctag, cvalue *C.yaml_char_t
 	var cimplicit C.int
 	var free func()

=== modified file 'encode_test.go'
--- encode_test.go	2011-11-24 19:47:20 +0000
+++ encode_test.go	2011-11-28 13:24:39 +0000
@@ -76,6 +76,10 @@
 		A int "a,omitempty"
 		B int "b,omitempty"
 	}{0, 0}},
+	{"{}\n", &struct {
+		A *struct{ X int } "a,omitempty"
+		B int              "b,omitempty"
+	}{nil, 0}},

 	// Flow flag
 	{"a: [1, 2]\n", &struct {

=== modified file 'goyaml.go'
--- goyaml.go	2011-11-24 19:47:20 +0000
+++ goyaml.go	2011-11-28 13:24:39 +0000
@@ -256,7 +256,7 @@
 	switch v.Kind() {
 	case reflect.String:
 		return len(v.String()) == 0
-	case reflect.Interface:
+	case reflect.Interface, reflect.Ptr:
 		return v.IsNil()
 	case reflect.Slice:
 		return v.Len() == 0

Along with EOF. This is useful for handling mixed diff and non-diff output. Note that "stray" content between diff files was already included in the extended headers for the next diff. All this commit does is return a partial FileDiff, with only extended headers and no actual diff content, along with an EOF.

Fixes https://github.com/sourcegraph/go-diff/issues/59

In some unified diff files there are such lines as Only in {path}: {filename}, meaning that certain files are only available in specific version. Those messages aren’t handled in any way, except when those lines are just catched in FileDiff.Extended for next FileDiff in the file. Are there any plans of handling these messages?

For example, those unique files can be picked up in next functions, representing them as FileDiff with only the value OrigName.

func (r *MultiFileDiffReader) ReadAllFiles() ([]*FileDiff, error)
func (r *MultiFileDiffReader) ReadFile() (*FileDiff, error)

Then for this file example, the output below for ReadAllFiles() would be expected. my_diff.txt

diff -u source_a/file_1.txt  source_b/file_1.txt
--- source_a/file_1.txt   2020-07-28 12:54:18.000000000 +0000
+++ source_b/file_1.txt  2020-07-28 12:54:18.000000000 +0000
@@ -2,3 +3,4 @@
 To be, or not to be, that is the question:
-Whether 'tis nobler in the mind to suffer
+The slings and arrows of outrageous fortune,
+Or to take arms against a sea of troubles
 And by opposing end them. To die—to sleep,
Only in source_a: file_2.txt
Only in source_b: file_3.txt

ReadAllFiles(myMultiFileDiffReader) -> {
    FileDiff{
        OrigName: “source_a/file_1.txt”, 
        OrigTime: ..., 
        NewName: “source_b/file_1.txt”, 
        NewTime: …,
        Entended: ..., 
        Hunks: …
    },
 
    FileDiff{
        OrigName: “source_a/file_2.txt”, 
        OrigTime: nil, 
        NewName: nil, 
        NewTime: nil
        …
    }, 

    FileDiff{
        OrigName: “source_b/file_3.txt”, 
        OrigTime: nil, 
        NewName: nil, 
        …
    }

}, nil

This package is currently accessible via sourcegraph.com/sourcegraph/go-diff but there isn't really a good reason for this, and we no longer serve vanity import paths at e.g. https://sourcegraph.com/sourcegraph/go-diff (the real location is now https://sourcegraph.com/github.com/sourcegraph/go-diff)

This package is in use by the Go build bots, it looks like: https://github.com/golang/build/blob/master/go.mod#L74

Some firewalls block Cloudflare, which Sourcegraph.com is hosted through -- making this package hard to fetch needlessly. Reported on Slack here: https://gophers.slack.com/archives/C9BMAAFFB/p1542362482359900

A PR for this would be very much appreciated :)

When running "go get -u" for github.com/johnwyles/vrddt-reboot:

sourcegraph.com/sourcegraph/[email protected]: parsing go.mod: unexpected module path "github.com/sourcegraph/go-diff"

~~DO NOT MERGE, this is a WIP PR and there are known TODOs to resolve. Early review (comments, suggestions) is welcome.~~ Edit: Ready for review and merge.

There are situations where a diff is empty. For example, if a blank file is added a removed. Or if a file is renamed without any changes. This change detects and handles those situations properly. Edit: Or if the file modified is binary.

Improve test coverage significantly.

TODO

• [x] Update .proto and regenerate .pb.go.

Fixes #10.

I've recently started using this library, to redact diffs that pertain to secret files from logs. It would be great if (Multi)FileDiffReader could yield content that is not part of a valid diff, so that the caller has the choice of whether or not to print it.

For example, the program colordiff will successfully print non-diff output, allowing it to be used to process input that is mixed diff and non-diff content.

What do you think of this feature, and would you be interested in a PR that implements it?

Unless I'm missing something, it seems the multi-file diff parser skips over changes to binary files (/cc @gbbr). Consider the following multi-file diff:

diff --git a/README.md b/README.md
index 7b73e04..36cde13 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,8 @@
 Conception-go [![Build Status](https://travis-ci.org/shurcooL/Conception-go.svg?branch=master)](https://travis-ci.org/shurcooL/Conception-go)
 =============

+This is a change.
+
 This is a work in progress Go implementation of [Conception](https://github.com/shurcooL/Conception#demonstration).

 Conception is an experimental project. It's a platform for researching software development tools and techniques. It is driven by a set of guiding principles. Conception-go targets Go development.
diff --git a/data/Font.png b/data/Font.png
index 17a971d..599f8dd 100644
Binary files a/data/Font.png and b/data/Font.png differ
diff --git a/main.go b/main.go
index 1aced1e..98a982e 100644
--- a/main.go
+++ b/main.go
@@ -6710,6 +6710,8 @@ func init() {
 }

 func main() {
+   // Another plain text change.
+
    //defer profile.Start(profile.CPUProfile).Stop()
    //defer profile.Start(profile.MemProfile).Stop()

There are 3 files changed. Two text files (README.md and main.go), one binary file (data/Font.png).

However, if I execute diff.ParseMultiFileDiff on that diff, the output is a slice with just 2 *diff.FileDiff entries, for the two text files. The binary file change is not there.

Reproduce sample

package main

import (
    "fmt"
    "strings"

    "sourcegraph.com/sourcegraph/go-diff/diff"
)

// fileDiffName returns the name of a FileDiff.
func fileDiffName(fileDiff *diff.FileDiff) string {
    var origName, newName string
    if strings.HasPrefix(fileDiff.OrigName, "a/") {
        origName = fileDiff.OrigName[2:]
    }
    if strings.HasPrefix(fileDiff.NewName, "b/") {
        newName = fileDiff.NewName[2:]
    }
    switch {
    case origName != "" && newName != "" && origName == newName: // Modified.
        return newName
    case origName != "" && newName != "" && origName != newName: // Renamed.
        return origName + " -> " + newName
    case origName == "" && newName != "": // Added.
        return newName
    case origName != "" && newName == "": // Removed.
        return "~~" + origName + "~~"
    default:
        panic("unexpected, no names")
    }
}

func main() {
    b := []byte(input)
    var o string
    if fileDiffs, err := diff.ParseMultiFileDiff(b); err == nil {
        for _, fileDiff := range fileDiffs {
            o += "\n" + "## " + fileDiffName(fileDiff) + "\n"
            /*o += "\n```diff\n"
            if hunks, err := diff.PrintHunks(fileDiff.Hunks); err == nil {
                o += string(hunks)
            }
            o += "```\n"*/
        }
    } else {
        o += "\n```\n" + err.Error() + "\n```\n"
    }
    fmt.Println(o)
}

const input = `diff --git a/README.md b/README.md
index 7b73e04..36cde13 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,8 @@
 Conception-go [![Build Status](https://travis-ci.org/shurcooL/Conception-go.svg?branch=master)](https://travis-ci.org/shurcooL/Conception-go)
 =============

+This is a change.
+
 This is a work in progress Go implementation of [Conception](https://github.com/shurcooL/Conception#demonstration).

 Conception is an experimental project. It's a platform for researching software development tools and techniques. It is driven by a set of guiding principles. Conception-go targets Go development.
diff --git a/data/Font.png b/data/Font.png
index 17a971d..599f8dd 100644
Binary files a/data/Font.png and b/data/Font.png differ
diff --git a/main.go b/main.go
index 1aced1e..98a982e 100644
--- a/main.go
+++ b/main.go
@@ -6710,6 +6710,8 @@ func init() {
 }

 func main() {
+   // Another plain text change.
+
    //defer profile.Start(profile.CPUProfile).Stop()
    //defer profile.Start(profile.MemProfile).Stop()

`

Apple's diff (at least the one shipping with macOS Ventura) does not include the fractional seconds or Timezone in its output.

This adds support for that so diffutils does not need to be installed.

I first encountered this problem using golangci-lint on macOS after upgrading to macOS Ventura.

Binary diffs are parsed correctly after #46, but we should still add support to get the decoded body, and a boolean flag to determine a file is binary.

Adds steps to the build pipeline (using GitHub Actions) to generate and upload LSIF for fast and precise code navigation on Sourcegraph. Uses the recommended lsif.yml file in Sourcegraph documentation.

This might actually be a bug in go, but I'm reporting it here anyway, since this breaks things here for now.

Recently, this library movied from sourcegraph to github. However, using the old sourcegraph.com name is now broken:

$ go get -d sourcegraph.com/sourcegraph/go-diff
go: sourcegraph.com/sourcegraph/[email protected]: parsing go.mod: unexpected module path "github.com/sourcegraph/go-diff"
go: error loading module requirements

The obvious fix is to just use the new github.com import path, but this is not always trivial. In particular, when go-diff is pulled in by a dependency that has the old url, running go get -u to update dependencies fails with the same error.

I guess that the problem is that go does not gracefully support updating to a new version in this case and updating across a rename (so 0.5.0 -> 0.5.1) needs changing of the import path. I can't see an easy way to fix this on the go side of things, so perhaps the fix here is to have sourcegraph.com not publish the 0.5.1 release (only 0.5.0), though that would also mean that people might never realize they're using an outdated version...

(also, I couldn't actually find the code on sourcegraph.com anymore, http://sourcegraph.com/sourcegraph/go-diff gives a 404, but apparently go get still manages to access the files)

NOTE: This is just a demo PR. It is not intended to be merged.

Makes it easier to canonicalize the output of PrintMultiFileDiff by adding a new PrintOptions arg that lets the caller specify that the diffs should be sorted before printing.