Your team is tasked with extending and rewriting the arithmetics library
holy-grail. You think the problem can probably be handled by your two lead
programmers sir-robin and sir-lancelot. They go to work immediately.
For the sake of this tutorial, you will wear both heads: that of lancelot and
robin who will work on different branches. For this, you'll first need to
create two directories which we'll call ./lancelot/ and ./robin. Next,
Please fork the project into your private repository. Let's say your
github account is called camelot. I have made this repo public, so that you
are able to fork it seamlessly. To fork, go to the repo
page and hit fork in the top task
list on the right. For the rest of the tutorial you'll work on your private
fork, where you can commit to master without disturbing the others. If you run
into bad trouble during the tutorial, just delete your private fork and fork
the BEL-CO one again.
Lancelot uses the command git clone https://github.com/camelot/github-tutorial.git to get a local copy of the
repository on his local computer. In the terminal, he runs
> pytest
executing all the four tests in the ./tests repository and receives the
following error message:
elements = ['a', 'b', 'c'], thesum = 'abc'
@pytest.mark.parametrize("elements,thesum", [
([1,2,3], 6),
(['a', 'b', 'c'], 'abc')
])
def test_sum(elements, thesum):
> assert arithmetics.sum(elements) == thesum
E TypeError: unsupported operand type(s) for +: 'int' and 'str'
tests/test\_adding.py:17: TypeError
================ 1 failed, 3 passed in 0.26 seconds =================
Immediately lancelot goes to the issues page of the github repository and creates a new issue:
sumfails when adding strings.It seems that builtin
sumdoes not work well with strings, however, all other tests seem to run through smoothly.
This instantiates Issue #1.
sir-lancelot creates a new branch which he calls fix-lancelot. He issues the command
> git checkout -b fix-lancelot
The command puts him on a new branch which is an exact copy of the current
master branch. He rewrites ./holyGrail/adding.py which then reads
> cat ./holyGrail/adding.py
def sum(s):
if len(s) == 0: return None
ans = s[0]
for si in s[1:]:
ans += si
return ans
def mean(s):
return sum(s)/len(s)
He adds all changes, commits, and pushes them with
> git add ./holyGrail/adding.py```
> git commit -m "Fixed sum."
> git push --set-upstream origin fix-lancelot
Note that git commit adds the commit permanently to his local branch,
but only git push will update (and even create) the remote branch with
the same name. The remote url id origin references the remote url, and
tells git about the repository where the new branch shall be created and
updated.
Content with himself, lancelot goes to lunch.
Having gone to the same ivy-league school as lancelot, Sir Robin takes similar initial steps
> git clone https://github.com/camelot/github-tutorial.git
> pytest
> git checkout -b fix-robin
He then goes on to do the exact same changes as lancelot, but he also reruns the test leading him to the following discovery:
elements = array([100.5 , 99.9 , 100.6 , 101.5 , 99.75, 99.75, 101.56, 100.75,
99.5 , 100.56, 99.56, 99.56, 100.25, ...5, 100. , 100.2 , 97.94,
99.75, 99.3 , 99. , 99.75, 101.8 , 100.6 , 99.44, 100.56],
dtype=float16)
themean = 100.0, tol = 0.1
@pytest.mark.parametrize("elements,themean,tol", [
([1, 2], 1.5, 1e-6),
((1e2+1.0*np.random.randn(1000)).astype(np.float16), 1e2, 0.1)
])
def test_mean(elements, themean, tol):
ans = holyGrail.mean(elements)
> assert np.abs(ans-themean) < tol, ans
E AssertionError: inf
E assert inf < 0.1
E + where inf = <ufunc 'absolute'>((inf - 100.0))
E + where <ufunc 'absolute'> = np.abs
tests/test\_holyGrail.py:25: AssertionError
========================================= warnings summary ==========================================
tests/test\_holyGrail.py::test\_mean[elements1-100.0-0.1]
/home/jus/Code/belco/github-tutorial/tests/../holyGrail/\_\_init\_\_.py:8: RuntimeWarning: overflow encountered in half_scalars
ans += si
-- Docs: http://doc.pytest.org/en/latest/warnings.html
========================== 1 failed, 3 passed, 1 warnings in 0.05 seconds ===========================
Sir Robin is furious, and bravely skips lunch to churn out the following code which now passes all tests:
> cat ./holyGrail/adding.py
import numpy as np
def sum(s):
if len(s) == 0: return None
ans = s[0]
for si in s[1:]:
ans += si
return ans
def mean(s):
return sum(np.asarray(s, dtype=np.float64))/len(s)
> git add ./holyGrail/adding.py
> git commit -m "Fixed sum and mean."
> git push --set-upstream master fix-robin
He checks everything three times and merges the changes into the master branch (on the github interface). Exhausted, Sir Robin falls asleep. Note that he could have done
> git checkout master
> git merge fix-robin
> git push
Sir Lancelot is back from lunch, runs pytest and discovers the failing mean
function that he had missed before. Immediately he goes to work, but he's
quite a bit smarter than robin. Instead of introducing a new dependency on
numpy he writes
> cat ./holyGrail/adding.py
builtinSum = sum
def sum(s):
if len(s) == 0: return None
ans = s[0]
for si in s[1:]:
ans += si
return ans
def mean(s):
return builtinSum(s)/len(s)
> git add ./holyGrail/adding.py
> git commit -m "Fixed mean."
> git push --set-upstream master fix-lancelot
Lancelot noticed that there are new commits in origin/master. He prays to
god that there's no conflict, but we all know there is; there always is. Sir
Lancelot checks diff of his own branch after updating his local master:
> git checkout master
> git pull
> git checkout fix-lancelot
> git diff master
diff --git a/holyGrail/adding.py b/holyGrail/adding.py
index 0af146f..f85057b 100644
--- a/holyGrail/adding.py
+++ b/holyGrail/adding.py
@@ -1,5 +1,5 @@
-import numpy as np
+builtinSum = sum
def sum(s):
if len(s) == 0: return None
@@ -9,4 +9,4 @@ def sum(s):
return ans
def mean(s):
- return sum(np.asarray(s, dtype=np.float64))/len(s)
+ return builtinSum(s)/len(s)
He doesn't know what all of this means, but it seems like there's now a conflict between his branch and the current master. He will need to resolve the conflict.
Sir Lancelot knows that conflict resolution is hard if you have lots of commits that need to be replayed on top of the new master. Therefore, he first cleans up his own branch by squashing the last two commits into one. He does:
> git rebase -i HEAD~2
# An editor opens ..
1 pick 9666fbe Fix sum.
2 pick 7ebc6dd Fix mean.
3
# .. lancelot changes it to ..
1 pick 9666fbe Fix sum.
2 squash 7ebc6dd Fix mean.
3
# .. saves and quits.
# Another editor opens ..
1 # This is a combination of 2 commits.
2 # The first commit's message is:
3
4 Fix sum.
5
6 # This is the 2nd commit message:
7
8 Fix mean.
9
# .. lancelot changes it to ..
1 Fixed without numpy.
2
# .. saves and quits.
> git log
commit 46d5b55424798a7b0f7991af3ac19da775bb3972
Author: lancelot <lancelot@camelot.uk.co>
Date: Fri Sep 21 05:12:52 2018 -0700
Fix without numpy.
He notice that the two commits have been replaced by a single one.
Running git status he recieves
On branch fix-lancelot
Your branch and 'origin/fix-lancelot' have diverged,
and have 1 and 2 different commits each, respectively.
(use "git pull" to merge the remote branch into yours)
nothing to commit, working directory clean
This is a dangerous state of divergence.
Sir Lancelot knows there's no turning back now. Bravely he enters
> git rebase master
First, rewinding head to replay your work on top of it...
Applying: Fix without numpy.
Using index info to reconstruct a base tree...
M holyGrail/adding.py
Falling back to patching base and 3-way merge...
Auto-merging holyGrail/adding.py
CONFLICT (content): Merge conflict in holyGrail/adding.py
error: Failed to merge in the changes.
Patch failed at 0001 Fix without numpy.
The copy of the patch that failed is found in: .git/rebase-apply/patch
When you have resolved this problem, run "git rebase --continue".
If you prefer to skip this patch, run "git rebase --skip" instead.
To check out the original branch and stop rebasing, run "git rebase --abort".
> cat holyGrail/adding.py
<<<<<<< 1dbed3799ec3fc2b0eee1bc7c2355ce3c8802816
import numpy as np
=======
ssum = sum
>>>>>>> Fix without numpy.
def sum(s):
if len(s) == 0: return None
ans = s[0]
for si in s[1:]:
ans += si
return ans
def mean(s):
<<<<<<< 1dbed3799ec3fc2b0eee1bc7c2355ce3c8802816
return sum(np.asarray(s, dtype=np.float64))/len(s)
=======
return ssum(s)/len(s)
>>>>>>> Fix without numpy.
The file holyGrail/adding.py has additional lines that are not python code.
<<< ... indicate conflicting parts in Sir Robin's remote commit that has been
played onto the file, ===== the end of that commit's changes and the
beginning of how the file logs in lancelot's branch, and >>>> Fix... the end
of that. Lancelot just deletes everything between <<< and >>>, so that
only his own code is left. Lancelot tries to continue:
> git rebase --continue
Applying: Fix without numpy.
This could have been way more complicated without the squashing in the beginning.
First lancelot pushes his diverged branch forcefully -f onto his
remote branch origin/fix-lancelot.
> git push -f
Online, the network now shows a linear history with his branch being one step ahead of master. Perfect. Finally he can merge his branch:
> git checkout master
> git merge fix-lancelot
> git push
THE END!