Stream: contributing

Topic: Strengths of each primitive/builtin Roc type

jan kili (Dec 14 2024 at 19:20):

I got these definitions and strengths from Roc's docs. Please suggest more, and we can compile a little guide to data modeling!

Bool

ex: Bool.true

• represent true or false

Dec

ex: 12.34

• represent a base-10 number with a fractional component
• represent currency

Dict

ex: Dict.fromList [("Fred", "my friend"), ("Ang", "a stranger")])

• represent an association between keys and values

F*

ex: 12.34f64

• represent a base-2 number with a fractional component
• represent graphical coordinates

I*

ex: -1_234i16

• represent an integer that can be negative

List

ex: ["my", "friend", "Fred"]

• represent a collection of similar values

Num

ex: 1234

• represent a number that could be either an Int or a Frac

record

ex: { my: Friend, name: "Fred" }

• represent a group of named values

Result

ex: Result Friend Stranger

• represent the result of an operation that could fail

Set

ex: Set.fromList ["my", "friend", "my", "Fred"]

• represent a collection of unique values, without any ordering

Str

ex: "my friend Fred"

• represent text

tag / closed/open tag union

ex: MyFriend "Fred"

• represent a custom value/set/subset... of... (we don't really provide a definition in Roc's docs)
• represent a custom set of possible variants (source: Elm's docs [1] [2])

tuple

ex: (my, Friend, "Fred")

• represent a group of unnamed values

U*

ex: 1_234u16

• represent an integer that cannot be negative

Anything I post below is just my opinion. :blush:

jan kili (Dec 14 2024 at 19:20):

Some strengths of tuples from the "Do we need tuples?" thread:

• tup.6 (source: Brendan)
• returning just two things that aren't really related, but I created as part of this function (source: Anthony)
• in when statements to pattern match on multiple things at once (source: Luke)

jan kili (Dec 14 2024 at 19:26):

U8:

• represent a byte

jan kili (Dec 14 2024 at 19:28):

U8:

• represent a bit if you're doing math with it

Bool:

• represent a bit if you're not doing math with it

jan kili (Dec 14 2024 at 19:32):

Dict:

• graduate from record if it (a) gets too many keys or (b) you start wanting tools like .values

jan kili (Dec 14 2024 at 19:35):

open tag union:

• simplify a module that (a) only handles a subset of your domain and (b) performs no-ops on outside variants

jan kili (Dec 14 2024 at 19:36):

(btw I'm very open to being wrong about these strengths I see - let me know if so!)

jan kili (Dec 14 2024 at 19:37):

record:

• graduate from tuple if it (a) gets too many fields or (b) touches an API

Notification Bot (Dec 14 2024 at 19:42):

This topic was moved here from #beginners > Strengths of each primitive/builtin Roc type by JanCVanB.

jan kili (Dec 14 2024 at 19:47):

Result:

• upgrade quick-and-dirty print-based debugging blocks like if foo bar then dbg bar to larger diversions like when bar is Err e -> with additional graceful handling of unmet expectations at runtime, without any obligation to treat every Err like an exit-triggering Fatal Error in the stronger conceptual sense

Brendan Hansknecht (Dec 14 2024 at 20:06):

Note, dict and set are kinda ordered

Brendan Hansknecht (Dec 14 2024 at 20:08):

Bool:

• Should almost never be stored. Essentially always better to store a tag. Really just exist for use in if conditionals.

Dict:

• For any sort of kv mapping where all keys have the same type and all values have the same type. Needed for any sort of dynamic mapping. Records are static.

Brendan Hansknecht (Dec 14 2024 at 20:11):

Dec:

• Recommended default fractional numeric type. Recommended because it is base 10 and behaves closer to what users expect. Use a float if perf ends up being an issue.

Brendan Hansknecht (Dec 14 2024 at 20:20):

Example of where dec would shine and float can suck: https://www.youtube.com/watch?v=SBfR3ftM1cU&t=3276s

Essentially 1597.0000 and 1597.0001 round to the same value in floats, but in dec, what you see is what you get. So thos are two distinct values.

jan kili (Dec 14 2024 at 20:24):

Brendan Hansknecht said:

Note, dict and set are kinda ordered

Should we edit its docs?
Screenshot_20241214_132349.png

Brendan Hansknecht (Dec 14 2024 at 20:28):

probably. Though maybe it is better to hide the ordering cause it is a kinda strange ordering

Brendan Hansknecht (Dec 14 2024 at 20:28):

basically, inserts are ordered

Brendan Hansknecht (Dec 14 2024 at 20:28):

removal swaps an item with then end and removes from the end

Brendan Hansknecht (Dec 14 2024 at 20:29):

so with a set if you insert 1, 2, 3, 4 then remove 1, you will get 4, 2, 3. Same with a dict just with attached values.

Richard Feldman (Dec 14 2024 at 21:07):

oh I think we should document the insertion order!

Brendan Hansknecht (Dec 14 2024 at 21:21):

Yeah, I guess my question is should we put it up front or as a detail later down

Brendan Hansknecht (Dec 14 2024 at 21:22):

I think for most people, the order shouldn't matter and it is more a distraction/implementation detail

Richard Feldman (Dec 14 2024 at 21:27):

ah, so I think that's how it ended up being for us, but it actually makes it useful in more situations so I think it's worth noting up front

Richard Feldman (Dec 14 2024 at 21:27):

just as a like "you can use it in these use cases" kind of thing

Brendan Hansknecht (Dec 14 2024 at 21:43):

:thumbs_up:

Notification Bot (Dec 14 2024 at 21:49):

2 messages were moved from this topic to #ideas > remove if to promote tags by Brendan Hansknecht.

Last updated: Jul 26 2025 at 12:14 UTC