A few months ago, we set up a "SQL book club" at work and tackled The Art of PostgreSQL by Dimitri Fontaine as our first book. Here's what I thought about it.
Even though working through a technical book isn't all that much fun, I've been able to stick to a "one part per two weeks" schedule (mostly because of the peer pressure in the form of an upcoming book club meeting). The book clocks in at about 430 pages, spread over eight parts with 51 chapters total. Reading one part and following along with the examples took be about three hours or so per part; we started in December and finished last week.
I've definitely learned a lot. Enough for a separate post, actually; I've taken notes for all chapters and want to polish and publish them, if only as a resource for my future self.
The book's tag line is
Turn Thousands of Lines of Code into Simple Queries
and I'm definitely more aware of things we could do using Postgres instead of application code.
The book is pretty complete in the sense that it starts from queries that just read data, goes on to data modelling and ends up with data manipulation and Postgres extensions. It is not a book about learning to write queries, but working through all the examples certainly helps by way of exposure to presumably well-written queries.
Throughout the book, I had lots of small "a-ha!" moments when an example used a
handy string function or some other functionality I hadn't been aware of but
could immediately start using; the chapter about setting up psql inspired me
to work through the complete documentation for it and definitely improved my
efficiency.
The book is expensive! There are four editions:
- "Standard" for $49 with the electronic version (PDF, EPUB and MOBI) and all
the queries as
.sqlfiles - "Full" for $99, with additionally a DB dump to load
- "Dead Tree" for $129, with additionally the printed book
- "Enterprise" for $229, which is like the Full Edition, but for 15 people and including a Docker container (but see below)
I pre-ordered the Full Edition for $69, if I remember correctly, which is still pretty steep, and work purchased the Enterprise Edition for the book club.
The book and data download takes place on Thinkific, a platform that seems to be meant for online courses, but all the features just get in the way of quickly downloading the files: I have to "start" or "replay" a "course" to get them. And I swear it aggressively messed up my password multiple times.
I might be overly sensitive to typos, but if I pay a lot of money for a book, I kind of expect some level of quality control, and not the outlook to Part 3 to mention "Writting SQL queries". There's a noticeable typo every few pages, and in some places the sentences could have done with a little love from a copy editor, leaving a not-so-great impression of "self published" overall, which is at odds with the steep price.
The Full Edition comes with a data set and a script to load it all, but I definitely didn't manage to run that successfully. Loading individual files was hit and miss, and I spent more than enough time just loading the example data to follow along.
Some of the queries contain little bugs like referring to the wrong schema, but those were fixed easily most of the time.
For some examples, the author only shows how he loads the publicly available data using a Common Lisp script, and I didn't even try to get those examples running. Too bad, because the queries on that data were about the most common pub names near a city identified by an IP address, which is the best kind of query.
Part of the allure of the Enterprise Edition is that it comes with a Docker container; spin it up, and start playing around with the data immediately! Great, isn't it?
Except, this:
The book has been out for a few months now; I don't think the sales pitch for the Enterprise Edition should even mention the Docker container if quite clearly it's not there yet.
I usually read my e-books with the Google Play Books reader, but it would stubbornly refuse to recognize the file as a valid EPUB file. I emailed the author about this, but I didn't get an answer; I also tried the chat function on Thinkific, which could as well just not be there, as I didn't get any response on there either1. Booh.
This one's really just my personal preference. I can live with this statement:
we now have color screens and syntax highlighting and we don't write all-caps code anymore... not even in SQL.
But the indentation advice to "right align top-level SQL clauses"... I could just not warm up to that.
Here's how the author would write a simple query:
select name, milliseconds
from album
left join track using(albumid)
where albumid = 1
order by trackid;
The whole thing starts with whitespace, and only on the last line can we see
that it's because order by is longer than select! The vertical separation
is not obvious to me. And aligning the table names in the from/join part
makes it look like there was an accident involving the space bar. I would write
this like
SELECT name, milliseconds
FROM album
LEFT JOIN track USING (albumid)
WHERE albumid = 1
ORDER BY trackid;
But really, that's not a complaint about the book. I'm just saying that until a team had agreed on a style for SQL, I'd try and nudge them away from the first one.
Each part ends with an interview; I didn't get too much value out of them. Some feel a bit too much like ads for Postgres, and I think a book about Postgres probably wouldn't include an interview with somebody who's super critical of it.
I don't regret spending quite some time to work through this book; I've learned
a lot and will often come back to it to look up things I've forgotten the
specifics about but remember their existence (JOIN LATERAL, anybody?).
It's far from perfect, though, and I don't like being promised (and paying for) things I don't get, such as a working EPUB version or the mythical Docker container.
Let's call this 4 out of 5 for content, but 3 out of 5 taking price and non-delivered features into account.
Footnotes
-
It seems that the chat function has since been removed. Makes sense. ↩