Database Reference
In-Depth Information
and one that contains all of the other plays. This was done without saying
which plays are histories or even what history means.
Visualizing Data in BigQuery
One of the problems with Big Data is that it becomes difficult to build an
intuition about your data. If you have a thousand rows in a table, you can
scan it, see whether one column is always numeric, or whether there are
null
s, and if there are, figure out what the
null
s represent. When you have
a million rows, however, it is harder to get a feel for the data—you're not
likely to want to read through all the data.
Data visualization tools can help you make sense of your data; good ones
make it easy to “see” various aspects of data, interactively. If you want to
plot one column by another, it is usually quite trivial, or if you want to get
statistics about unique values, the tool usually computes them for you. Most
data visualization tools work great on your million row table; however, if you
have a billion rows, they may start to take minutes to compute simple values,
or hours to produce a graph.
There are a number of data visualization tools that combine with BigQuery
to get the performance you'd expect from a small dataset on your Big Data
tables of virtually any size. This section introduces two of them that take two
different approaches to data visualization: Tableau and Bime.
Visualizing Your BigQuery Data with Tableau
Tableau has quickly become the gold standard for data visualization; it
makes it easy to get a feel for your data in a way that is difficult when
you're just writing queries. You can drag and drop fields to plot one against
another, apply aggregations, and so on. It has a lot of automatic intelligence
built in so that it can pick the best type of graph for your data.
Tableau's main visualization product, Tableau Desktop, is a thick client. This
means that a lot of the visualization work is being done on your machine,
rather than in the cloud. At a time when everything seems to be moving
to the cloud, this feels like a bit of a throwback. However, there are good
reasons to do work on the client: It enables you to build much snappier
visualizations because you don't have to wait until some remote server
decides to process your request.
