I understand it as an attempt to get very basic, manual syntax highlighting. If all you have is white text on black background, then I do see the value of making keywords easy to spot by putting them in all caps. And this probably made sense back when SQL was first developed, but it's 2023, any dev / data scientist not using a tool that gives you syntax highlighting seriously needs to get with the times
Partially, yes. I personally use an IDE with excellent syntax highlighting and those have been around for at least two decades. You are, however, often transplanting your SQL between a variety of environments and in some of those syntax highlighting is unavailable (for me at least) - the all caps does help in those rare situations.
More importantly though it helps clearly differentiate between those control keywords (which are universal) and data labels (which are specific to your business domain). If I'm consulting on a complex system that I only partially understand it's extremely helpful to be able to quickly identify data labels that I'm unfamiliar with to research.
Well then use all-caps keywords whenever working on those systems, I don't care. But an edge case like that shouldn't dictate the default for everyone else who doesn't have to work on that, that's all I'm saying.
There are several cases where you'll be limited to console only, or log files, or many many other situations. Good coding practices just makes life easier all around.
JetBrains IDEs - IntelliJ, WebStorm, PyCharm, GoLand, etc., all support highlighting SQL embedded in another source file or even inside markup files like YAML. Does your IDE not support this?
As the other commenter said, the Jetbrains IDEs do this perfectly fine. Although I'd also argue that if you're working with SQL from within another language already, a DSL wrapper is probably gonna be the better way to go about this.
Unfortunately RustRover is still garbage for actual usage. And I refuse to use an ORM when I can just write the SQL in a more common syntax that everyone understands across every language instead of whatever inefficient library-of-the-week there is. Raw SQL is fine and can be significantly more performant. Don't be scared.
I'm not talking full blown ORM here, not a fan of those either. I'm talking about some light weight wrapper that basically just assembles SQL statements for you, while giving you just a little more type safety and automatic protection against SQL injection, and not sacrificing any performance. I'm coming from the JVM world, where Jooq and Exposed are examples of that kind of thing.
I'm currently using SQLx which you write raw queries in and it validates them against a currently-running db, using the description of the tables to build the typing for the return type instead of relying on the user. It makes it pretty hard to write anything that supports injection
Happens at compile time! It's relatively quick. You can also run a command to write the query results to file for offline type checking which is mostly useful for CI
My ide isn't limited to color when it comes to highlighting, so being color blind generally shouldn't be a problem. Set keywords to underlined, bold, italic, whatever works for you.
Your other examples I can see, but at least at my work those are rare edge cases, and I'd rather optimize for the brunt of the work than for those. Of course at other places those might be much more of a concern.
Sorry, to clarify, not everything is in all caps. I'll append my prefered syntax below
WITH foo AS (
SELECT id, baz.binid
FROM
bar
JOIN baz
ON bar.id = baz.barid
)
SELECT bin.name, bin.id AS binid
FROM
foo
JOIN bin
foo.binid = bin.id
The above is some dirt simple SQL, when you get into report construction things get very complicated and it pays off to make sure the simple stuff is expressive.
I've seen both approaches and I think they're both quite reasonable. An indented join is my preference since it makes sub queries more logically indented... but our coding standards allow either approach. We've even got a few people that like
FROM foo
JOIN bar ON foo.id = bar.fooid
JOIN baz ON bar.id = baz.barid
I believe this has been proven. It's because capital letters all have the same shape whereas lower case letters do not. So your brain can take shortcuts to reading lower case but cannot with upper case.
Also most if not all editors will highlight SQL keywords so it's probably not too hard to discern SQL commands and everything else in modern day.
I'm quite aware... basically it means that novice devs can create a table in camelCase and query in camelCase... but you can clean it all up as long as they didn't realize you needed double quotes.
Fair point. I always disliked the design because ORMs pretty much always use quotes, so an entity-first approach can create a lot of tables with capital letters if you're not careful, which is then really annoying if you need to use raw SQL for anything.
It's an English literacy thing - we have several non-native English speakers and using only singular avoids making those folks' lives harder. Besides it's really nice to autopilot that categoryid is a foreign key to the category table. It also simplifies always plural words... I haven't yet written CREATE TABLE pants but if I ever do there's zero chance of me creating a pantid.
I tend to use underscores on join tables so table foo_bar would have a fooid and a barid. I have somewhat soured on this approach though since there are a lot of situations where you'll have two m-m relationships between the same two tables with a different meaning... and having a fixed formula for m-m tables can make things ugly.
If I get to design another greenfield database I'll probably prefer using underscores for word boundaries in long table names.