brettp
We ran into an issue today with structs that I found a bit surprising.
The essence of the issue is:
- A
structis aMapunder the hood - you can manipulate a
structwithMapfunctions, deleting keys that should be there and adding keys that shouldn’t - as long as the
__struct__key is still there, theMapis treated as astructincluding in Elixir pattern matching
Here’s some code which demonstrates this:
defmodule Sandbox do
@enforce_keys [:mandatory]
defstruct [:mandatory, :optional]
def struct_foo do
s1 = %Sandbox{mandatory: "must have this"}
do_something("s1", s1)
s2 = Map.delete(s1, :mandatory)
do_something("s2", s2)
s3 = Map.put(s1, :bogus, "blah")
do_something("s3", s3)
end
def do_something(s, %Sandbox{} = thing) do
IO.puts "#{s} #{inspect thing} is a %Sandbox"
end
def do_something(s, thing) do
IO.puts "#{s} #{inspect thing} is just a thing"
end
end
iex(24)> Sandbox.struct_foo()
s1 %Sandbox{mandatory: "must have this", optional: nil} is a %Sandbox
s2 %{__struct__: Sandbox, optional: nil} is a %Sandbox
s3 %{__struct__: Sandbox, bogus: "blah", mandatory: "must have this", optional: nil} is a %Sandbox
One interesting thing is that IO.inspect seems to know the difference between the original struct and one that’s been hacked into a Map, but Elixir pattern matching doesn’t.
So you almost certainly should not be manipulating a struct in this way, and you certainly shouldn’t be deleting keys from structs. The question is: should (could) Elixir do anything to stop you doing this? Should Map functions check for __struct__ and behave differently if they are asked to do something nefarious to a struct? Should a Map function which manipulates a struct at the very least also remove the __struct__ key so it returns a Map that is not treated as a struct any more?
Where Next?
Trending in Discussions
Other Trending Topics
Latest on Elixir Forum
Sponsor Spotlight
Elixir, Erlang & messaging conferences from Code Sync. Next up: ElixirConf US (10–11 Sep) & Code BEAM Europe (21–22 Oct), Haarlem + online.
Categories:
Sub Categories:
Forums
Popular Tags
- #ecto
- #liveview
- #troubleshooting
- #learning-elixir
- #library
- #deployment
- #erlang
- #testing
- #genserver
- #mix
- #absinthe
- #remote-other
- #otp
- #plug
- #how-to-question
- #macros
- #postgres
- #elixirconf
- #channels
- #exunit
- #discussion
- #code-sync
- #podcasts
- #javascript
- #onsite
- #dialyzer
- #docker
- #authentication
- #ai
- #umbrella
- #full-time-contract
- #podcasts-by-brainlid
- #ecto-query
- #blog-post
- #elixirconf-us
- #elixir-ls
- #phoenix_html
- #iex
- #graphql
- #genstage
- #websockets
- #supervisor
- #advent-of-code
- #distillery
- #processes
- #elixirconf-eu
- #api
- #forms
- #security
- #performance
Our Sponsors
Build Elixir applications with speed and confidence.
Supporting innovation across the BEAM ecosystem.
We build reliable cloud platforms for business-critical systems.
Catch errors, track performance, monitor hosts and more.
Error tracking, logs, uptime, and Just Enough APM. Built for developers.
Practical resources that improve the lives of professional developers.
Master modern tools, write cleaner code, and tackle real-world challenges.
Bringing learning, work and compliance together since 1987.
The team behind Membrane, Popcorn, LiveDebugger. Available for hire.
Courses that'll move you from confusion to "Aha, now I get it!"
Showing Posts 1 to 10- Show Best Posts
- Show All (oldest first)
- Show All (newest first)
LostKobrakai
If you know you’re working with a struct and want to preserve its integrity use
struct/2orstruct!/2.dimitarvp
That’s a really weird code though. It’s known that you shouldn’t try adding/deleting keys to/from a struct.
Why would you have code like your above?
brettp
Well I don’t have code like that – that’s just sample code to demonstrate the underlying behaviour: that Elixir allows you to manipulate a struct into something that looks nothing like the struct, but it still treats it as a struct.
(For the curious, the issue was that somewhere in a codebase someone had, with good intentions, deleted from a struct all keys whose value was
nil. This caused issues elsewhere in the codebase – specifically when relying on the default@deriveimplementation ofJason.Encoder– where code assumed that if something was a struct then all the keys should be there.)I know you shouldn’t write code like that; my question was whether Elixir should - or indeed could - be doing something to stop you.
dimitarvp
So if you are dealing with a larger code base and more developers then I’d say it makes sense to convert the strict structs to loose maps on their way out of the code you control. Let the other modules / projects do whatever with them and then try and convert them back to strict structs via the
structfunction on their way in back to your code.OvermindDL1
I’m surprised dialyzir didn’t catch it, this seems like a case it would catch?
jeremyjh
If the type is not opaque what could dialyzer catch here? It is not even aware of structs. Making a struct opaque is possible but then you have to write accessors for every piece of data you might want to use.
OvermindDL1
But dialyzer does know that access on a struct’s/map’s fields that don’t exist is an error.
jeremyjh
I don’t think any error will be raised in the above code. The Map.put option succeeds, so what is there for dialyzer to complain about? The elixir compiler does complain about access or pattern match of non-existing fields, but dialyzer will never even see that code. It should catch cases that must lead to an error in exactly the same circumstances that it would for a bare map, but no others unless user-defined types are supplied.
brettp
One of the things I find odd is that
IO.inspectsomehow knows the difference between a struct and theMapwith a__struct__key that is returned from e.g.Map.delete:struct:
%Sandbox{mandatory: "must have this", optional: nil} is a %Sandboxmap with a
__struct__key:%{__struct__: Sandbox, optional: nil} is a %Sandboxjosevalim
Correct.
It could but it would make all maps operations more expensive, which is mostly why we don’t. If we had a static type system, doing Map.put/3 on a struct would certainly fail.