Day 2
git-intro
Branches and merging
https://coderefinery.github.io/git-intro/06-branches/
- Is it better to use “git switch”?
git switch works with the newer version of Git.- We just don’t teach it yet, because it’s new enough it would cause confusion when it doesn’t work for some people.
- But if it is available on your version, I recommend to use it since there is less ambiguity than
git checkout (which can do many different things).
- can you show more than just the last command
- C-b ESC
<down> should increase the top one by five lines
- can you again explain quickly what git graph does
- It shows the full “commit graph”, how each commit relates to each other. Also the branches. It is very useful as a “overall summary” of your latest commits
- When I have unstaged changes at master and switching to master branch via
git checkout master, will the unstaged changes be lost?-
- It never loses changes (as far as I see). It will first try to move the changes to the new branch. If that doesn’t work, it shows a message saying “can’t change branches with uncommited changes” or similar.
- When this happens, I ues
git stash before to save changes, , do the switch, and then git stash pop to get it back. Don’t worry about this yet, though
- The output of my git command show in refreshed clean window , instead of show right after the command line. Is there a way to change this?
- This can be changed by changing the “pager” (see below)
- Which operating system are you using?
- A quick workaround is to pipe it to
cat, e.g. git show hash1 | cat
git --no-pager log stops the pager, too.
- this works for me, thanks!
- What does the git do with the different versions of files in different branches? Are they stored separately or only their differences?
- Git does not store differences but actual snapshot. But it does it very efficiently by chopping up files into portions, storing them in a tree structure and reusing common parts. When you run a
git diff, the differences are actually computed on the fly, they are not stored. This storage format works less well for large binary files.
- This goes into a bit more detail: https://coderefinery.github.io/git-intro/13-under-the-hood/
- What is the difference between git graph and git log –oneline? Does it matter which one to use?
- graph has
--decorate (show branches and tags) and --all (show other branches) and --graph (show how the commits connect), too. graph is just an alias we define, so you can use what you’d like.
- you will see the difference once we have more branches and tags
- If I committed to the wrong branch, can I easily move the commit to the correct branch?
- Yes, it’s “easy” but requires some care. See below. I recommend using
git graph between every stage, to ensure you understand all the steps. Then, it becomes natural.
- I would do this (let’s imagine I wanted to create
somebranch but forgot and accidentally committed to master):
- Create a branch referencing this new commit:
git branch somebranch (now somebranch points to this commit)
- Run
git graph to verify that the new branch really points to that new commit.
- Do a
git log or git graph to find out to which hash you want to move master back, say the hash is abc1234
- Make sure you are on
master
- Move
master back to abc1234: git reset --hard abc1234 (be careful, git reset --hard “throws away” commits but if you do this by mistake, you can recover from it with the help of git reflog)
- Why do we always need to “add” before “commit”? Is it ok to just “commit”?
- It lets you check that you’ve added all the right things first. You can use
git diff and git diff --staged to make sure all the right things are about to be committed.
- But you don’t have to, there are several options that directly commit without adding (as long as it’s added at least once)
- You can also do directly
git commit filename(s) but we recommend to do this in two steps since you can review your commit with git status
- If I clone a repo, why do I have to use git checkout branch_name to make the branch visible in git branch output?
- We will clarify this later but after
git clone you get all the remote branches, but git branch only shows you local branches (clone will create a local branch for you, typically master)
- To see all branches, included remote branches, use
git branch --all
git checkout somebranch will create a local branch referencing a remote branch somebranch, if the latter exists as a remote branch origin/somebranch
- When I change on some branch, eg less-salt, the changes are also visible in master. Does that make sense? Doesn’t seem like clean separate branches for me…
- It shouldn’t, we should investigate in the breakout room. What does
git graph show?
- It doesn’t show my branch even though it’s there. I can checkout to it.
- We should look at this in a breakout room in a bit. There are plenty of ways for stuff to get mixed up.
- What is your breakout room number?
-
- So it seems before staging and commiting any changes on branches are “global”. Only after staging there solely on the branch.
- “git branch” after the mergers still shows the other branches. So it is still possible to get back to those and rebranch?
- you can jump back to a branch that has already been merged and keep working on it if you want. But normally you organize your work in feature branches which get deleted after merging. Sometimes though you want to merge the other way, e.g. from
master to your feature branch (to keep up to date with the master branch). Then obviously you don’t want to delete master
- I got merge conflict, are we going to discuss this? x2
-Error/conflict: git merge less-salt
Auto-merging ingredients.txt
CONFLICT (content): Merge conflict in ingredients.txt
Automatic merge failed; fix conflicts and then commit the result. --- what to do?
- Yes, soon. If you get a merge conflict, you probably adjusted the same lines. It’s OK, there is a way to recover from this in the lesson.
- Try to follow along as best as you can, and we’ll
- please before merging a lot, could we discuss, what exactly happened with the files?
- Git is able to automatically combine modifications done on two branches done in the same file, unless both branches modified the same code portion.
- Then it creates a new commit, a merge commit.
- If the same code portion has been modified differently on two branches, and we try to merge these, Git issues a “conflict” and we need to resolve it, we will see later how.
-
- Merge operations can only happen between the master and a child branch?
- Merge always modifies the current branch you are on but you can merge any other branch into the current branch.
- We typically merge feature branches into
master once the feature is done.
- You sometimes want to merge
master to the feature branch to update it with features/bugfixes done on other branches.
- Why are merged branches still visible when using “git branch”? (continued)
- The
git merge does not delete the branch after merging. How can we delete it. Should we delete the branches?
- they are not deleted automatically.
- If you want to delete a branch after merging you have to do
git branch -d branch-name
- Is
Gitk any good for visualization? Do you use it?
- Personally I use
git graph and also I look at the “network” (via insights) on GitHub to browse branches.
- Yes, it’s good but for day to day work I find
git graph more convenient. Other things are nice for really complicated stuff.
- How does git select between the changed lines when merging? Child branch overrides the parent (experiment -> master)?
- By default Git is very conservative and will not automatically select. If both modified the same code portion, Git stops and asks you to resolve the conflict.
- If you are sure that you want to keep the one or the other without reviewing, you can instruct
git merge to automatically follow the one or the other if it hits a conflict (example later in the session).
- What happens if I make add-pepper and add-chili branches, commit a change into the ingredients.txt so that ‘1 chil1’ and ‘1 pepper’ are on ‘line 2’ on both files since the files are different in different branches. If I now merge them BOTH with the master, what happens?
- The first merge will work (does not matter which branch)
- But merging the second branch will issue a “conflict” where you need to decide whether to keep the chili or the pepper (we will later see how to resolve conflicts).
- So when I have git graph as following:
```
| * 2ae77cc (less-salt) reduce amount of salt
- | e77e41a Merge branch ‘experiment’
|\ \
| * | 86772b6 (experiment) reduced amount of cilantro to 1 tbsp
| * | 1686bed add cilantro to recipe
```
And I say ‘git show e77e41a’, I do not see the changes caused by merge.
- The merge commit doesn’t introduce any actual changes itself, so that is probably why it shows nothing.
- What do the colours in git graph mean?
- vertical lines: the colors are a visual aid to help you distinguish the different branches. Note that the asterisks symbolize commits
- What happens if I delete the last branch, or a branch that uniquely refers some commits? Are these commits then lost?
- git will raise a warning if you try to delete an unmerged branch with
git branch -d. If you’re really really sure that you still want to delete it you can override the warning with git branch -D
git branch -d experiment less-salt
error: The branch 'less-salt' is not fully merged.
If you are sure you want to delete it, run 'git branch -D less-salt'.
Deleted branch experiment (was d78c499).
- The merge probably has not worked somehow. We should examine in breakout rooms.
- Can one undelete a deleted branch?
- Yes, will tell more in a bit
- If you remember the hash, you can recreate it with
git branch somename somehash
- If you forgot the hash, you can for some time find it using
git reflog
- We had a conflict in our rebasing. I wanted to know how to open the conflict in the diff tool etc., like I usually do in VSCode. It would be nicer than deleting.
- I think
git difftool would do it. But, I haven’t done this too many times, so maybe someone else can comment
- If your breakout room had technical problems (no helper, not enough time, something else), please comment here so we can know the extent of the problems (and breakout room number):
- I again kept getting kicked out. Although happened less serverely than yesterday, breakout room 10.
- Can you try Zoom in Chrome?
- I’ll try from Chrome in the next break then.
Branching and conflicts
https://coderefinery.github.io/git-intro/08-conflicts/
- How
git branch -d works? should my HEAD be in the source branch? can I delete from a different branch a branch that it is merged in another? Clarification: You have master create from there branch A, add 1 commit, create from there branch B, add 1 commit, mergeB into A, move the HEAD to master. From there can you git branch -d B? or you have to be in A?
- Can you clarify some?
- You shouldn’t delete the currently checked out branch (I think it won’t let you)
- It removes the “pointer” to commits, so if those commits are merged somewhere else, all commits still exist. If those commits aren’t merged somewhere, they are “orphaned” and don’t appear anywhere (but not deleted right away)
- To clarification: yes, you could delete
B if I understand correctly. It seems like a reasonable thing to do in that case (of course depending on your goals).
- Is it possible to create a new branch with just few files. Let’s say if the master had thousands of files and I want to create a new branch but making a copy of the entire master will take up lot of space. In such a situation, could we copy just the files that will be worked on rather than the entire master?
- Note that creating new branches does not mean using up more space. So create many branches. Git will not duplicate files between branches, in Git branches are not implemented as copies. A branch in Git is a file containing 40 characters and all it contains is the hash that it points to. Branches in Git are like sticky notes.
- You can create unrelated, so called “orphan” branches which have no connection to the branch where you issue this command but I think what you wanted is still to have some connection but only with a subset of the files? Because with the orphaned branch you are completely disconnected.
- It is also very fast at switching branches: only modified files have to be adjusted.
- Can you comment on why the git graph looks that way after merging the branch from the exercise?
- why is there no ‘lines’ going to the right like with earlier branches?
- do you mean at the commits on top?
- yes, earlier branches got those lines going out to the right when something was changed, while the branch from the exercise did not have any in git graph.
- I made dislike branch from like branch. How can I move it to master?
- See two points below for the same question
-
Answer to a question regarding update file with a following checkout <branch>
This is explained in git help checkout (here is a text cut):
git checkout <branch>
To prepare for working on <branch>, switch to it by updating the index and the files in the working tree, and by pointing HEAD at the branch. Local modifications to the files in the working tree are kept, so that they can be committed to the <branch>`.
- What if you mistakenly branch from like-cilantro? Can you just delete it or can you rewind it back to the correct commit?
- if you haven’t committed to it yet, you can delete the branch with
git branch --delete somebranch, go to the right hash, and create it there
- alternative is to move it to the right place (
somehash) with git checkout somebranch followed by git reset --hard somehash
- before moving branches I would always first create a “backup” branch pointing to the same commit as the branch I am about to correct, before I move it. If I mess up, I can then always go back and still have it.
- Has anyone ever merged not with master?
- Git merge always changes current branch which I am on (check with
git graph) and I can merge any other branch into the current branch. My current branch does not have to be master.
master is not different/special, only difference is the name. One can even delete the master branch.- Does this answer?
Breakout room task:
Sharing repositories online
https://coderefinery.github.io/git-intro/09-remotes/
This is just a preview of tomorrow, don’t worry too much about full details here!
- Can I create a private repository instead?
- You always can. in this case, you can if you want. It won’t affect much for this particular exercise.
Exercise: this type-along as an exercise:
https://coderefinery.github.io/git-intro/09-remotes/#type-along-create-a-new-repository-on-github
- How do we delete a remote repository name when it was mistyped in the terminal (after creating the repo in github)? Thanks.
git remote remove <name>git remote lists the current remote names.
History inspection
https://coderefinery.github.io/git-intro/10-archaeology/
- Is there a way to completely delete a file from history? Say, I push a very large file accidentally, and I don’t want it in my repository.
git filter-branch can do this. Search the web for descriptions of this.
- How does
git grep differ from grep -R?
- Quite similar,
git grep has more options for searching history, only added files, etc.
- Any particular reason why the rvest repository is in a detached HEAD state when you clone it?
- I’m not sure why it would be - can you show us?
- I think I understood why, it’s because it’s cloning from a branch “git clone –branch v0.3.5 https://github.com/tidyverse/rvest.git”. In “git clone help” says that “–branch” “can also take tags and detaches the HEAD at that commit in the resulting repository.”
- Please leave angle brackets in backquotes, otherwise coloring for the whole document gets off
Exercise:
Main goal: https://coderefinery.github.io/git-intro/10-archaeology/#archaeology-exercise
Optional: https://coderefinery.github.io/git-intro/10-archaeology/#git-bisect-exercise
15 minutes, ending at xx:55
git checkout <hash> <branch-name> creates a new branch, would a soft reset also be technically correct?
- No, this is not correct. HEAD will move to the
<hash>. All commits following the <hash> will be lost, or almost lost. There is ways to get it back, but a git graph will not show the commmits.
- Is it a very bad habit to use
git commit -a in personal work (automatic staging of modified files)? Or should this be avoided for some reason?
- If you are OK with having everything in giant commits, it’s OK
- In the future, when you come back, you wont’t b able to tell things apart much. If you only care about the big picture, it’s OK
- Another way to look at it, it’s better than nothing! (I have an alias that does this, but don’t use it very often)
- I have problems with the command ‘git commit -p’. This is the output in Win 10/anaconda prompt: git: ‘add–interactive’ is not a git command. See ‘git –help’. fatal: interactive add failed
- Interesting. Does it say
add--interactive without a space after add?
- It says “git: ‘add–interactive’ is not a git command. See ‘git –help’.” So it seems that there is no space after add
- It looks like
git commit -p is trying to internally call git add-interactive instead of git add --interactive (the right command, with a space). So… I’m not sure how to debug this.
- As a workaround, I would use
git add -p to add parts interactively (you can select which chunks to add), and then git commit which will commit those added changes. I often do this anyway, when I want to make sure I got everything.
Feedback, day 2
One thing that is good, one thing that should be improved: