All right, let's do a little bit of a Cloud Datalab demo. First thing you might notice is that in the products and services menu, when you scroll down and click on BigQuery or click on Dataprep, that's typically how you've accessed those different big data resources. For a Cloud Datalab, it's actually a notebook software, much like iPython or Jupyter notebooks that we talked about a little bit earlier, and you need to actually load that onto a machine. Now normally, statisticians or data scientist would load that on their local laptops, and then sample data on their own hard drives and perform those machine learning or statistical model calculations. But, a little bit better than performing on your local machine, is actually performing on a virtual machine. Getting access to a lot more resources and computing power. So to spin up a virtual machine, click on Google Cloud Shell in the upper right hand corner, and Google Cloud Shell is that terminal view or we can spin up and spin down resources. In order to activate resources, we are going gcloud compute and see what zones are actually available for our virtual machine to be spun up in, and since I'm in the US, I'm gonna grab US central. And to invoke datalab, to create a brand new datalab is just a simple as writing datalab create. If I can spell insights, data and demo insights virtual machine, sure and then with a flag, let's specify the zone of US Central one dash f. As you see, this points can take about 5-10 minutes to spin up the virtual machine and then load it on the Cloud Datalab, and then from within there, in actually a separate browser window as we're going to show you in just a minute, you can then have access. Almost like you would locally in your local host to that virtual machine which has datalab on it. And we're back. Once you get a message that looks like this, click on the web preview button, and change your port, that's how you know your local host on your virtual machine is ready to view. We'll change our port here to 8081 and hopefully, boom. Welcome to the Google Cloud Datalab. So the unit of work that we're going to be doing here is not going to be a SQL file or text file for all our codes. Actually, it's going to be the individual notebook. So, you can organize multiple notebooks and folders. And again, if you're already familiar with iPython notebook it's the dot iPython notebook format file extension. We are going to create a brand new notebooks. So we can just do a quick demo. As we covered in a little bit of the lecture slides, notebooks primarily have two types of components that are inside of them. One is the code, which we're going to run, and it can be things other than BigQuery code, which is pretty cool. So things like accessing machine learning models, building visualizations and charts, Python code, Pandas DataFrames, all that kind of good stuff. And then, my personal favorite is you can annotate or add Markdown if you've used like GitHub before or if you've used iPython notebooks before. This can be simple things like headers, this is high. This is a header, and I'm writing this inside of what's called the notebook cells. This is the actual cell. Multiple cells can be found in a notebook. And then if I actually ran this, that's a code section. So if I ran this instead, boom, there is a header. Now this looks like to be code, so we are going to write some code in there. And I want my code to appear after my Markdown comments. And, I actually want to do a sample SQL. Am also going to do IRS and see if we can edit this, top 10 US states with the most charities. And you can give a little bit of reference. Great. And again, you can run individual cells, you can run more than one cell at once, and boom. Now we have some very quick Markdown for our notebooks. And the great thing about this is again, if you want to share this notebook with somebody else, they'll be able to see all of your comments, all of your different cells for the codes, they can copy the notebook and then make edits and changes and run it themselves. So it's a little bit more interactive than just say sharing a BigQuery script with somebody, because it gives a lot more additional context. First and foremost, we need some BigQuery code. In order to do that, we can invoke BigQuery directly within the Cloud Datalab cells. That's one of the benefits of using Google Cloud Datalab is the interface with a lot of the Google Cloud Platform tools like BigQuery, TensorFlow, and a bunch of other big data tools. So we need to specify this is going to be a BigQuery query. Let's see if I can do this from memory. We're going to do, select the cells to count(*), and I'm doing this by hand but of course, you'd have a BigQuery opened in a separate tab and then just copy in your SQL script that you have there. I want the count of charities, I want the state I need to pull it from, this is always the hard part of BigQuery. Public data is the projects, IRS 990 is the data set, IRS 990 the organizational detail tables of the EIN is where we run report from. And much like if you remember for doing aggregate function like count star, we need to group everything else, group by state, and I am going to order it by highest number of charities first, and I only want the top 10. So I will throw in 11 under there. Okay great. Hopefully, our SQL syntax is correct, and let's go and run this cell and see what we return. So its actually going to pass it through to BigQuery. Hopefully, no errors. And then boom, immediately back, we get this pretty results over here. California has 163 thousand charities, and you see just how fast that was interfacing directly with BigQuery as well. All right, but you're saying to yourself, "I could just do that in the BigQuery user interface." It's great that you can add these comments in hyperlinks and Markdown there, but so far I haven't seen anything that's wowed me. Let's just keep going. So many other great things that you can do, is you can do visualizations directly inside of Cloud Datalab notebook. So I am going to copy this query, in much like you've seen with temporary tables or permanent tables, you can actually store the results of that query into a name. I want to name the results of this query to be charities by state, and interesting, right? If I actually just added that flag and call it charities by state and I hit run, you won't actually see the results, but they're actually being stored behind the scenes. And the cool part, what that allows us to do as we keep adding more and more code, is invoke some really easy to call charting mechanics. So I'm going to say, hey build me a chart for the data. I want you to actually use this thing that I just stored for you. I named it charities by state. And then since you're doing a column visualization, a column chart, these are the fields that I want you to use. I want you to use state on the x-axis and on the y-axis, I want the number of charities for that state. And let's see what that looks like. Bam. So immediately, you can start to see how dangerous Cloud Datalab notebook can be. It combines a little bit of what you've seen with Google Data Studio and building up these quick visualizations. Also at BigQuery, you can see the individual queries that made up the results here, and also with a little bit of markdown documentation so that you can add a lot more context and helpful notes for your other fellow data scientists or data analysts that will be looking at your results. This is just scratching the surface. This is just using BigQuery interfacing through Cloud Datalab. But you can imagine doing a lot of your pre-processing on BigQuery, and then invoking something like if you're familiar with Pandas DataFrames or TensorFlow APIs. This is the exact same type of work that you can be doing within here if you know a little bit of a Python or R, that will allow you to access those machines on the models and those machines on the APIs. And once you've gotten to the end here, again keep in mind that notebook is kind of like an artifact, so you can save it, give it a name. We'll call it irs_demo_top_charities_by_state, save the notebook. You can clone it, you can have different versions. It's going to be version control just by checking it into like a Git repository. And of course, you can just download the notebook as an artifact and then share it with others if you're not using something like version control. Remember that's highly recommended. And then you can share it or just if you already have a notebook that's copied in, you can fork a copy of somebody else's, that notebook that's been created. So it's very, very collaborative. So the keyword that you remember for all of these Cloud Datalab demos and notebooks, is that they're meant to be collaborative, they're meant to be editable by your peers. Okay, That's a very, very high level view of Cloud Datalab. Let's head back into the lecture for a summary.
