feat(hogql): stringOr ^^

[mariusandra]

Problem

We want to easily get the first not empty string in Hog/QL:

In JavaScript:

const a = null
const b = ''
const c = 'banana'
const d = 'dingo'

console.log(a || b || c || d) // 'banana' -- this is what we want
console.log(a ?? b ?? c ?? d) // ''

In Hog:

let a := null
let b := ''
let c := 'banana'
let d := 'dingo'

print(a || b || c || d) // 'bananadingo' -- double pipe is string concatenation
print(a ?? b ?? c ?? d) // ''
print(a or b or c or d) // true

In HogQL

select null || '' || 'banana' || 'dingo' // 'bananadingo' -- double pipe is string concatenation
select null ?? '' ?? 'banana' ?? 'dingo' // ''
select null or '' or 'banana' or 'dingo' // errors
select coalesce(null, '', 'banana', 'dingo') // ''

The only real way to get banana is:

let a := null
let b := ''
let c := 'banana'
let d := 'dingo'

print(notEmpty(a) ? a : notEmpty(b) ? b : notEmpty(c) ? c : d) // 'banana'

... but we could do better.

Changes

Introducing stringOr(null, '', 'banana', 'dingo') and its shorthand ^^

Why ^^?

Looking at my keyboard, I found these underutilised symbols in our language: ! $ % ^ & _ ~. I excluded symbols that are common bitwise operators or used elsewhere in Hog, and added a bunch of doubles.

The question became: which of these is the most natural/obvious?

print(null ! '' ! 'banana' ! 'dingo')    // feels like "not"
print(null @ '' @ 'banana' @ 'dingo')    // banana at dingo? doesn't read right
print(null $ '' $ 'banana' $ 'dingo')    // somehow this makes me think of joining strings
print(null _ '' _ 'banana' _ 'dingo')    // this too
print(null !! '' !! 'banana' !! 'dingo') // this is "truthiness" in a lot of places
print(null $$ '' $$ 'banana' $$ 'dingo') // feels heavy, and also like concat
print(null %% '' %% 'banana' %% 'dingo') // feels like weird string formatting
print(null ^^ '' ^^ 'banana' ^^ 'dingo') // ^ is XOR, so it is a funky "or" and I can kind of feel it
print(null && '' && 'banana' && 'dingo') // we don't use it, but that's definitely not an "or"
print(null ~~ '' ~~ 'banana' ~~ 'dingo') // it looks like joining with a rope
print(null __ '' __ 'banana' __ 'dingo') // feels like joining/underscoring strings

... and that's how I settled on ^^

How did you test this code?

I didn't yet. This is a WIP in the spirit of "the fastest way to get the right answer is to post the wrong answer on the internet".

To test locally, run pip install ./hogql_parser. If you change the ^^ in HogQLLexer.g4 and want to play around, run pnpm run grammar:build followed by pip install ./hogql_parser.

Alternatives

We claim back ||, however it is a standard SQL concatenation operator. We have advertised it and people are using it in queries, so it'll be a hard fight.

[posthog-bot]

It looks like the code of hogql-parser has changed since last push, but its version stayed the same at 1.0.45. 👀
Make sure to resolve this in hogql_parser/setup.py before merging!