Diff for "TutorialFirstChange"

Tutorial - making our first change

Carrying forward from TutorialHistory, we are inside our my-hello repository that we cloned in TutorialClone.

It is good Mercurial development practice to isolate each change in a separate [:Repository:repository] (see also ["WorkingPractices"]). This prevents unrelated code from getting mixed up, and makes it easier to test individual chunks of work one by one. Let's start out by following that model.

Our silly goal is to get the "hello, world" program to print another line of output. First, we create a new repository called my-hello-new-output, by cloning from my-hello, for our little project.

$ cd ..
$ hg clone my-hello my-hello-new-output

In this case, the clone command prints no output if it succeeds.

LyleJohnson adds: For my installation of Mercurial 0.9.4 on Mac OS X (from MacPorts), I did get one line of output, as follows.

DanielLagesse adds: I also see this. Mercurial 0.9.4 on Ubuntu 7.10.

$ hg clone my-hello my-hello-new-output
2 files updated, 0 files merged, 0 files removed, 0 files unresolved

Note: Notice that we have given our new repository a descriptive name, basically identifying the purpose of the repository. Since making a [:Clone:clone] of a repository in Mercurial is cheap, we will quickly accumulate many slightly different repositories. If we do not give these repositories descriptive names, we will rapidly lose the ability to tell them apart.

Now it's time to make a change in the new repository. Let's go into the [:WorkingDirectory:working directory], which is simply our name for the directory where all the files are, and modify the source code with our favorite editor:

$ cd my-hello-new-output
$ vi hello.c

The contents of hello.c initially look like this:

/*
 * hello.c
 *
 * Placed in the public domain by Bryan O'Sullivan
 *
 * This program is not covered by patents in the United States or other
 * countries.
 */

#include <stdio.h>

int main(int argc, char **argv)
{
        printf("hello, world!\n");
        return 0;
}

Let's edit main so that it prints an extra line of output:

(...)

int main(int argc, char **argv)
{
        printf("hello, world!\n");
        printf("sure am glad I'm using Mercurial!\n");
        return 0;
}

Once we're done, we quit out of our favorite editor, and we're done. That's it. The edit is now ready for us to create a [:ChangeSet:changeset].

But what if we're interrupted, and we've forgotten what changes are going to make it into the changeset once we create it? For this, we use the status command.

$ hg status
M hello.c

The output is terse, but prefix M is simply telling us that hello.c has been modified, so our change is ready to go into a changeset.

We may also examine the actual changes we have made to the file using the diff command:

$ hg diff
diff -r 82e55d328c8c hello.c
--- a/hello.c   Fri Aug 26 01:21:28 2005 -0700
+++ b/hello.c   Fri Sep 30 10:27:47 2005 +0800
@@ -12,5 +12,6 @@
 int main(int argc, char **argv)
 {
        printf("hello, world!\n");
+       printf("sure am glad I'm using Mercurial!\n");
        return 0;
 }

In case we wish to discard our changes and start over, we may use the revert command to restore hello.c to its unmodified state (or use the -a option to revert all files). Just make sure you know this is what you really want.

$ hg revert hello.c

KenMin adds: revert command also renames the modified file hello.c to hello.c.orig

$ hg status
? hello.c.orig

If we change our mind again and want to reuse the modification we have made, we just remove the unmodified state of hello.c and rename the modified hello.c.orig to hello.c

$ rm hello.c
$ mv hello.c.orig hello.c
$ hg status
M hello.c

The act of creating a changeset is called [:Commit:committing] it. We perform a commit using the commit command:

$ hg commit

This drops us into an editor, and presents us with a few cryptic lines of text.

Note: The default editor is vi. This can be changed using the EDITOR or ["HGEDITOR"] environment variables. Also, the [:Manifest:manifest] hash might be different, depending on how you typed and saved the file.

(empty line)
HG: manifest hash 14595beb70bcfb74bf227437d70c38878421c944
HG: changed hello.c

The first line is empty and the lines that follow identify the files that will go into this changeset.

To commit the changeset, we must describe the reason for it (see ["ChangeSetComments"]). Let's type something like this:

Express great joy at existence of Mercurial
HG: manifest hash 14595beb70bcfb74bf227437d70c38878421c944
HG: changed hello.c

Next, we save the text and quit the editor, and, if all goes well, the commit command will exit and prints no output. If you quit the editor without saving the text, commit will abort the operation, so you may change your mind before committing.

Note: Before committing anything into a serious project, you may want to configure a meaningful username in ~/.hgrc; see ["QuickStart2"].

Let's see what the status command will tell us now?

$ hg status

Nothing! Our change has been committed to a changeset, so there's no modified file in need of a commit. Our [:Tip:tip] now matches our working directory contents.

We can now examine the change history for our new work:

$ hg log
changeset:   2:a58809af174d
tag:         tip
user:        mpm@selenic.com
date:        Fri Aug 26 01:26:28 2005 -0700
summary:     Express great joy at existence of Mercurial

(...)

There it is! We've committed a changeset.

Note: The user, date, and [:ChangeSetID:changeset ID] will of course vary.

As we discussed in TutorialClone, the new changeset only exists in this repository. This is a critical part of the way Mercurial works.

To share changes, we must continue to TutorialShareChange.

CategoryTutorial