Re: [git-users] Git Puzzle: Managing multiple websites based on the same code
On Tue, Jun 18, 2013 at 2:42 PM, Dyske Suematsu wrote: > Currently I manage them with SVN. The original, generic site is in the > "trunk" and for each new client, I create a branch. If I make a change to > the trunk that needs to be propagated to all the branches/sites, I just run > svn merge to merge trunk into branches. I've been doing this for some years > and it's been working fine. Do the same thing in git, works fine, its just you can't do a single clone, and get the trunk and all branches and everything checked out at the same time. Clone multipe times: mkdir SITE cd SITE for every customer A, B, C git clone REPO.git A&& cd A && git checkout origin/A now you have basically the same thing. You could do this with different github projects, if for some reason you wanted your customers to see only a single git repository named after them, but behind the scenes you would make them all be branchs of the same code base. The way to do it is to NOT use github's fork button. Instead, have a main repo, MAIN.git, then make a new one for each customer A.git, B.git, etc. You would then manage each independently, using cmd line tools, instead of the github web UI, pushing and pulling and merging between them. > Another suggestion, which is obvious but would not work, is to create a > branch for each website. This wouldn't work because the concept of "branch" > in Git is almost completely different from SVN. Git does not create a > separate directory for each branch; it just switches the same directory. So, > if you have 20 websites, you would have to constantly switch between > different websites and you could never pull up two sites side-by-side on > your local machine. Furthermore, websites these days have files and folders For some reason you seem to be stuck on the notion you can only do a single git clone on your machine, and you have to do all your work in that single clone... this isn't even true with svn, I usually have a svn checkout from root that I sporadically keep up to date, but then in my working directory I have multiple checkouts of my projects trunk, under different names,for different bits of work I'm doing. So just clone the repo multiple times. You can even recreate the svn directory structure locally if it makes you feel good: mkdir -p REPO REPO/branches cd REPO; git clone repo.git -t master cd branches; for A, B, C git clone ...; git checkout branch ... > Submodules and Subtree are two other suggestions but these would require me Stay away from them. -- You received this message because you are subscribed to the Google Groups "Git for human beings" group. To unsubscribe from this group and stop receiving emails from it, send an email to git-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
[git-users] Git Puzzle: Managing multiple websites based on the same code
Hi All, I spent a whole week researching this and found no answer. So, I'm hoping that someone here could help me. Here's my situation which I don't think is so unique or strange. I have a generic website that I always use as my starting point to build websites for all of my clients. Since each client's needs are different, I end up editing different parts of the original code. (In other words, I cannot separate and isolate files that all sites share in common.) As you could imagine, without some sort of version control, it would be very difficult to manage all the websites. The same fix/change would have to be applied to all the sites individually. Currently I manage them with SVN. The original, generic site is in the "trunk" and for each new client, I create a branch. If I make a change to the trunk that needs to be propagated to all the branches/sites, I just run svn merge to merge trunk into branches. I've been doing this for some years and it's been working fine. Now, I would like to switch to Git because it's more powerful overall, but I cannot figure out how I can manage these sites with Git. I searched this Google Group and found someone who asked a similar question. I also asked some of my programmer friends who have been using Git for a long time. They all suggested to create a fork for each website. I thought it would work, so I started implementing this but hit a wall. The issue is that on Github, each fork requires a unique user. Say, I create the "Official" repo under my "Organization" account. The URL would be: https://github.com/mycompany/basesite I then create a fork of this, and I would end up with: https://github.com/username/basesite Now, if I want to create another site, I would have to create a new user because each fork is differentiated only by the user name. If I have to create a new user for each website, eventually, I'll end up with a whole bunch of user accounts to manage. And, since this must be a private repo, each user would cost $7 a month and it can quickly amount to thousands of dollars a year. So, this strategy hit a wall here. Another suggestion, which is obvious but would not work, is to create a branch for each website. This wouldn't work because the concept of "branch" in Git is almost completely different from SVN. Git does not create a separate directory for each branch; it just switches the same directory. So, if you have 20 websites, you would have to constantly switch between different websites and you could never pull up two sites side-by-side on your local machine. Furthermore, websites these days have files and folders that the users upload (like photos). You wouldn't want to put them into the repo but you do need them to switch when you change from one website to next. Managing all this would be confusing and difficult under the same working directory. Submodules and Subtree are two other suggestions but these would require me to separate and isolate the files that are common to all sites. As I said above, I cannot do this because each client has different needs. I don't think my situation is so uncommon. It's a situation where you spin off a bunch of different versions from the same original codebase and each is a legitimate product of its own (not a temporary state to be discarded or merged at some point). Git is a powerful and flexible tool, so I'm thinking that there HAS TO BE a solution for this. If anyone has any idea, I would appreciate it very much (I would buy you lunch if you are in New York). Thank you, Dyske -- You received this message because you are subscribed to the Google Groups "Git for human beings" group. To unsubscribe from this group and stop receiving emails from it, send an email to git-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.