Who has used version control? CVS? Subversion?
Has anyone used distributed version control? Git? Mercurial? Bazaar? Darcs?
So you're writing a Super Awesome web app. You write your code on your laptop, test it, and when it works, you copy it up to the live server. Everybody is happy.
Then one day you copy up a change, and the live site breaks. Something on the server is different (different version of Python? Key dependency not installed? File permissions problem?) and code that works locally doesn't work there. You frantically try to debug it quick, no luck. This is going to take some time to solve. Meanwhile, your site is down and all your potential visitors are getting big nasty errors. You really, really wish you could just quickly rollback to the last version that worked. But oops, you just copied over it. Sorry.
Or, your site gets popular and you get a friend to help you with coding. They've got a copy of the site, and you've got a copy of the site. You make some changes, they make some changes. You copy your version to the server, they copy their version to the server. And - they just overwrote your changes. Oops.
u r doin it wrong
Version control lets you track changes to your project, instantly restore any past version of your project, and merge changes with other people working on the project simultaneously.
Version control is one of the two most important software development tools, after the IDE or code editor. There is no reason to ever work on a software project without using version control.
It's also really good for non-programming tasks, like tracking versions of a static HTML web site, or server configuration files, or any text files. I even use version control on my todo lists :-)
Mercurial is already installed on the lab Macs. To create a
Mercurial repository (a directory of version-tracked
files), just type hg init myrepo. (Hg
is the symbol for the element mercury). This will create the
directory myrepo, which will be empty except for a
hidden directory .hg (the initial dot makes a file or
directory hidden in Unix, but you can see it with ls -a
myrepo/.) This .hg directory is where
Mercurial will store all the historical version data about your
files. Otherwise, the myrepo directory is empty and
ready for you to add some files to it.
You've probably had enough Python for one night, so we'll start by
just tracking some text files. cd myrepo/,
then emacs file1.txt and put a couple lines of text in
it, whatever you want. Save the file and C-z to put emacs in the
background.
hg status will summarize the current status of your
repository. At the moment, it will tell you:
? file1.txt
which means it doesn't know what that file is. To tell it to start
tracking that file, do hg add file1.txt. Then hg
status will tell you:
A file1.txt
which means file1.txt is ready to be added.
At this point, Mercurial still hasn't committed any changes to your project's version history. It's still just telling you what it's ready to commit.
Before we actually commit, let's tell Mercurial what editor you
prefer for entering your commit log messages. The
default on these Macs is set to vi, which is fine for
Paul Ortman, but the rest of you might want to run export
EDITOR=emacs.
Now hg commit to commit your current changes (the
addition of file1.txt) as a new version (the first version!) of your
repo. Mercurial will bring up Emacs for you to enter a log message
detailing what you did in this changeset. Enter something like "added
file1.txt", save, and exit.
Now hg status won't show you anything! This means that
your files are exactly the same as the most recent committed version:
because you just committed! If hg status doesn't list
anything, that means you have no outstanding changes waiting to be
committed.
Run hg log. It will show you the commit you just made, with the commit message:
changeset: 0:15747f0e0cb3 tag: tip user: carljm@un004-imac25.goshen.edu date: Thu Sep 10 19:07:44 2009 -0400 summary: added initial file
Now let's bring up Emacs again (fg) and make some
changes to file1.txt. Save your changes and C-z to get
back to the shell. Run hg status - it will tell you
M file1.txt, meaning that file1.txt has been
modified.
hg status also probably lists ?
file1.txt~; it doesn't know what to do with the backup file
that Emacs automatically creates. Let's tell Mercurial to ignore all
of those backup files. fg back into Emacs, C-x
f and open a new file .hgignore. Put these lines
into that file:
syntax: glob *~
Now hg status won't show the backup file; it will list
? .hgignore, so run hg add .hgignore to add
your .hgignore file to the project. Then run hg commit -m 'added
.hgignore file' .hgignore.
This time it won't bring up Emacs, because you supplied the
-m flag with a commit message, right on the command
line. This is what you'll probably do most of the time; it's quicker
than going into Emacs for every commit message.
Also, you gave a file name, .hgignore, on the command line. This
tells Mercurial to only commit the changes to that file. If you run
hg status again, you'll see that file1.txt
is still listed as modified; the changes to it were not committed.
Run hg diff, and Mercurial will show you the full
changes that are ready to be committed:
diff -r b5a4ca908b84 file1.txt --- a/file1.txt Thu Sep 10 19:33:09 2009 -0400 +++ b/file1.txt Thu Sep 10 19:33:18 2009 -0400 @@ -1,3 +1,3 @@ Here is a text file. -With some text in it. +With some changed text in it.
Looks ok, so hg commit -m 'frobjammered the contrapulator'. hg log will now show three commits.
Chapters 1 and 2 in the Definitive Guide to Mercurial.
Nothing complicated: just put your Python homework for this week inside a Mercurial repository directory. I'd like to see several commits in your log at sensible versions (i.e. each time you reach a point where something is working) showing your progress as you moved through the assignments.