Gnafu Braindumps

TL;DR

 git bisect start <bad_commit> <good_commit>

Git bisect can save you a lot of time if you have to track the exact commit that introduced that bug (you may also find out that you introduced that yourself).

Let’s say you have a repo in a good state.

As time goes on, other developers commit their code, make changes, and everything look right because conflicts are resolved, tests pass, builds build.
So far so good.

And then QA hits: that thing you knew was working 100% it isn’t anymore.

You get a glass of liquor, sit by the fireplace and comand your butler “Release the hounds“. And start the git blame.

But scanning through a week worth of work, if the repo is large and the developers are quite a few, can take quite some time to find the culprit, expecially if you have to run some tests to check wether the commit is good or not.

Here git bisect comes handy.

With git bisect you just have to identify one bad commit and a good commit, then it will get the range between them and checkout in the middle (hence bisect).

You test the checkout and tell git if it is good or bad, and it will repeat the process so the range becomes shorter and shorter, untill a single commit is left.
That’s where the bug was introduced.

Let’s see an example:

The current commit is known to be bad
The know good commit is aa238a0

> git bisect start      // start the bisect
> git bisect bad       // mark the current commit at bad
> git checkout aa238a0 // go to the good commit
> git bisect good     // mark the current commit good

Git will checkout a commit in the middle so you can review it.

> git bisect good
Bisecting: 5 revisions left to test after this (roughly 3 steps)
[f9c99aef47dd865a4bd15545a18de7b03f09d780] hey oh

in_progress

Review the files, they are still bad, mark the commit again.

> git bisect bad
Bisecting: 2 revisions left to test after this (roughly 2 steps)
[6055332a77d6925b9225159a80623cf6d10be97b] another test

in_progress_2

This one is good, mark it.

> git bisect good
Bisecting: 0 revisions left to test after this (roughly 1 step)
[826696448e48ad9ed52239a3e746616e6eeb1989] teipo

Bad again, mark it.

> git bisect bad
Bisecting: 0 revisions left to test after this (roughly 0 steps)
[74ab31899d3ddf3aa5d70e4bb318280eb89fe1c3] typo

Good, mark it.

> git bisect good
826696448e48ad9ed52239a3e746616e6eeb1989 is the first bad commit
commit 826696448e48ad9ed52239a3e746616e6eeb1989
Author: it_was_indeed_yourself
Date: Mon Jun 5 23:39:26 2017 +0200

teipo

:000000 100644 0000000000000000000000000000000000000000 2011421b6a38b0fd51a2e600de4ef5fe6e26496f A bad-file.png
:100644 100644 21941c42e8565b36db082cdeca95e3072392b748 6b51d1612fab6f3137d41f760d0edd1524543068 M oh.txt

Here we go! we found the first bad commit!

in_progress_4

This is obviously a very simple example.
For more complex codebases git even supports scripts to be run automatically to decide whether a commit is good or bad.

See more about this time-saving tool in the official documentation.

DigitalGlobe Maps API had a free plan to use their maps, but it will expire on May 31 2017.

UPDATE: It’s expired now 🙂

So here’s a quick comparison I did with their maps.

DigitalGlobe distributes their maps via MapBox so the API is the same: you just have to use different map names and API keys.

At low zoom levels (1:144000 and smaller scale) the maps are the same, it’s the standard MapBox Satellite map.

After zooming in we start to notice the differences.
On the left you can see the DigitalGlobe Recent imagery, on the right MapBox Satellite one.

The recent one looks like a patchwork because it actually is: every time they acquire a new image from the satellites they update the map.
At first it can look a little ugly compared to the other one, but there is a reason for that: every processing on the images to make them “look good” will just degrade the quality of the image. If you want recent images it because you want to know what’s there, not just a good looking background.

Let’s have a closer look.

In this comparison we can see a new mall in the center. That mall is not visible in MapBox Satellite map (right), that means that those images dates back at least to 2013, since the construction started in 2014. DigitalGlobe Recent on the other side (left) shows the mall and two other buildings on the side, dating the images near the end of 2015.

Two years ahead in this city, but I’m sure more interesting ones are updated more frequently.

They also have transparent maps with road names in both English (where available) and Local language.

It was interesting playing around during the free trial, I think I’m going to keep an eye on that maps for future projects.

The GIS viewer in the screenshots is MapStore2 , a free open-source map editor made by my company GeoSolutions. Feel free to give it a try!

Gnafu Braindumps

Git bisect saves the day

TL;DR

Quick, before it expires!