Squash commits into one with Git

In Git you can merge several commits into one with the powerful interactive rebase. It's a handy tool I use quite often; I usually tidy up my working space by grouping together several small intermediate commits into a single lump to push upstream.

Step 1: choose your starting commit

The first thing to do is to invoke git to start an interactive rebase session:

git rebase --interactive HEAD~N

Or, shorter:

git rebase -i HEAD~N

where N is the number of commits you want to join, starting from the most recent one. For example, this is a hypothetical list of commits taken from the git log command, while I'm working on a generic feature Z:

871adf OK, feature Z is fully implemented      --- newer commit
0c3317 Whoops, not yet...
87871a I'm ready!
643d0e Code cleanup
afb581 Fix this and that
4e9baa Cool implementation
d94e78 Prepare the workbench for feature Z
6394dc Feature Y                               --- older commit

And this is what I would like to do:

871adf OK, feature Z is fully implemented      --- newer commit --┐
0c3317 Whoops, not yet...                                         |
87871a I'm ready!                                                 |
643d0e Code cleanup                                               |-- Join these into one
afb581 Fix this and that                                          |
4e9baa Cool implementation                                        |
d94e78 Prepare the workbench for feature Z     -------------------┘
6394dc Feature Y                               --- older commit

Obtaining:

84d1f8 Feature Z                               --- newer commit (result of rebase)
6394dc Feature Y                               --- older commit

Notice how a rebase generates a new commit with a new hash (84d1f8 in the example above). So in this case the command would be:

git rebase --interactive HEAD~[7]

because I want to combine the last seven commits into one, and d94e78 Prepare the workbench for feature Z is the seventh one.

I have tons of commits to squash, do I have to count them one by one?

A downside of the git rebase --interactive HEAD~[N] command is that you have to guess the exact number of commits, by counting them one by one. Luckily, there is another way:

git rebase --interactive [commit-hash]

Where [commit-hash] is the hash of the commit just before the first one you want to rewrite from. So in my example the command would be:

git rebase --interactive 6394dc

Where 6394dc is Feature Y. You can read the whole thing as:

Merge all my commits on top of commit [commit-hash].

Step 2: picking and squashing

At this point your editor of choice will pop up, showing the list of commits you want to merge. Note that it might be confusing at first, since they are displayed in a reverse order, where the older commit is on top. I've added --- older commit and --- newer commit to make it clear, you won't find those notes in the editor.

pick d94e78 Prepare the workbench for feature Z     --- older commit
pick 4e9baa Cool implementation 
pick afb581 Fix this and that  
pick 643d0e Code cleanup
pick 87871a I'm ready! 
pick 0c3317 Whoops, not yet... 
pick 871adf OK, feature Z is fully implemented      --- newer commit

[...]

Below the commit list there is a short comment (omitted in my example) which outlines all the operations available. You can do many smart tricks during an interactive rebase, let's stick with the basics for now though. Our task here is to mark all the commits as squashable, except the first/older one: it will be used as a starting point.

You mark a commit as squashable by changing the work pick into squash next to it (or s for brevity, as stated in the comments). The result would be:

pick d94e78 Prepare the workbench for feature Z     --- older commit
s 4e9baa Cool implementation 
s afb581 Fix this and that  
s 643d0e Code cleanup
s 87871a I'm ready! 
s 0c3317 Whoops, not yet... 
s 871adf OK, feature Z is fully implemented      --- newer commit

[...]

Save the file and close the editor.

Copy Data Wizard in Azure Data Factory

In this example, we are going to copy the themes.csv file from Rebrickable into a blob container called lego in our Azure Data Lake Storage Gen2 account.

From the Azure Data Factory Home page, click copy data:

This opens the Copy Data Wizard. Let’s walk through each step!

1. Properties

On the Properties page, give the pipeline a name and description. Keep the default “run once now” option:

I chose the name Lego_HTTP_to_ADLS_Themes. This tells me that I’m copying data from a website (HTTP) into Azure Data Lake Storage Gen2 (ADLS). This name makes sense in my head, but you need to find a naming convention that works for you :)

Click next to move on to the Source properties.

2. Source

On the Source page, we will first create a new linked service to Rebrickable, then create a new dataset to represent the themes.csv file.

Click create new connection:

Search and select the HTTP Linked Service:

Give the linked service a name and description, and use the base URL cdn.rebrickable.com/media/downloads/. (You can find this URL by inspecting the links on rebrickable.com/downloads. Keep the last slash.) Change authentication type to anonymous. Click create:

I chose the name HTTP_Lego, because I like to prefix my linked services with the connection type.

The linked service has now been created, yay! Make sure it’s selected and click next to move on to the dataset properties:

Since we specified the base URL in the Linked Service, we only have to specify the file name themes.csv.gz in the relative URL. Keep the other default options. Click next:

This next part feels kind of like magic, especially if you have been working with SQL Server Integration Services (SSIS) in the past. The Copy Data Wizard now inspects the file and tries to figure out the file format for us. But… since we are working with a gzipped file, it doesn’t make a whole lot of sense yet…

Let’s fix that! Change the compression type to gzip. Tadaaa! Magic! Without us doing anything else manually, the copy data wizard unzips the CSV file for us and shows us a preview of the content:

If you are working with a raw CSV file, the copy data wizard can detect the file format, the delimiter, and even that we have headers in the first row. But since we are working with a gzipped file, we have to configure these settings manually. Choose first row as header:

If the headers are not detected correctly on the first attempt, try clicking detect text format again:

You can now preview the schema inside the gzipped file. Beautiful! :D

Click next to move on to the Destination properties.

3. Destination

On the Destination page, we will first create a new linked service to our Azure Data Lake Storage Gen2 account, then create a new dataset to represent the themes.csv file in the destination.

Click create new connection:

Select the Azure Data Lake Storage Gen2 linked service:

Give the linked service a name and description. Select your storage account name from the dropdown list. Test the connection. Click create:

I chose the name ADLS_cathrinewadls, because I like to prefix my linked services with the connection type.

The second linked service has now been created, yay! Make sure it’s selected, and click next to move on to the dataset properties:

Specify lego as the folder path, and themes.csv as the file name. Keep the other default options. Click next:

Enable add header to file and keep the other default options:

Click next to move on to the Settings.

4. Settings

On the Settings page, we will configure the fault tolerance settings. This is another part that feels like magic. By changing a setting, we can enable automatic handling and logging of rows with errors. Whaaat! :D In SQL Server Integration Services (SSIS), this had to be handled manually. In Azure Data Factory, you literally just enable it and specify the settings. MAGIC! :D

Change the fault tolerance settings to skip and log incompatible rows:

At this time, error logging can only be done to Azure Blob Storage. Aha! So that’s why we created two storage accounts earlier ;) Click new:

The Copy Data Wizard is even smart enough to figure out that it needs to create an Azure Blob Storage connection. Good Copy Data Wizard :D Give the linked service a name and description. Select your storage account name from the dropdown list. Test the connection. Click create:

I chose the name ABLB_cathrinewblob, because I like to prefix my linked services with the connection type.

Specify lego/errors/themes as the folder path:

Click next to move on to the Summary.

5. Summary

On the Summary page, you will see a pretty graphic illustrating that you are copying data from an HTTP source to an Azure Data Lake Storage Gen2 destination:

Click next to move on to Deployment.

6. Deployment

The final step, Deployment, will create the datasets and pipeline. Since we chose the “run once now” setting in the Properties step, the pipeline will be executed immediately after deployment:

Once the deployment is complete, we can open the pipeline on the Author page, or view the execution on the Monitor page. Click monitor:

Success! ✔🥳 Our pipeline executed successfully.

We can now open Azure Storage Explorer and verify that the file has been copied from Rebrickable:

Summary

In this post, The Copy Data Wizard created all the factory resources for us: one pipeline with a copy data activity, two datasets, and two linked services. This guided experience is a great way to get started with Azure Data Factory.

Next, we will go through each of these factory resources in more detail, and look at how to create them from the Author page instead of through the Copy Data Wizard. First, let’s look at pipelines!