# Dictionary showing invalid keys
**URL:** https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422
**Category:** General Usage
**Tags:** question
**Created:** [March 21, 2023, 10:25pm UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422 "2023-03-21T22:25:38Z")
**Posts on this page:** 16
**Page:** 1
### Author:  [@gxyau](https://discourse.julialang.org/u/gxyau)
#### Post date: [March 21, 2023, 10:25pm UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/1 "2023-03-21T22:25:38Z")
Hello there,
Been playing around with Julia a bit just to practice some topics. I have recently started using `Graphs.jl` and found some unexpected behaviour. A MWE and the output has been included below.
Long story short, I created a graph with some edges, and I would like to associate a probability to each edge, which is constant. For each edge, I cast the `Graphs.SimpleGraphs.SimpleEdge{Int64}` to a tuple as the key of the edge probability, whose keys are tuples and values are its probability. However, when checking the key of the dictionary, it shows that there are 16 keys instead of 3 which I was expecting. Any idea what happened?
Thank you all very much in advance!
```plaintext
# MWE
using Distributions, Graphs, Random
G₀ = SimpleGraph(3, 0)
V₀, E₀ = vertices(G₀), edges(G₀)
add_edge!(G₀, 1, 2)
add_edge!(G₀, 2, 3)
add_edge!(G₀, 1, 3)
# Defining probability of connection and probability of disconnection
probs = rand(Uniform(0, 1), G₀.ne)
const edge_prob = Dict{Tuple{Int64, Int64}, Float64}(Tuple.(E₀) .=> probs)
println("The dictionary has length $(length(probs))")
println("The keys has length $(length(edge_prob.keys))")
```
```plaintext
# Output
The dictionary has length 3
The keys has length 16
```
---
### Author:  [@odow](https://discourse.julialang.org/u/odow)
#### Post date: [March 21, 2023, 10:42pm UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/2 "2023-03-21T22:42:03Z")
You want `keys(edge_prob)`, not `edge_prob.keys`.
---
### Author:  [@Benny](https://discourse.julialang.org/u/Benny)
#### Post date: [March 21, 2023, 11:05pm UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/3 "2023-03-21T23:05:12Z")
The `keys` method gets you an object that holds the keys. The `.keys` field is an `Array` holding the keys and _empty slots_. I think 16 entries is as low as it ever gets, and more space is allocated to make sure the dictionary isn’t too full. IIRC, hash table performance generally degrades fast beyond 2/3 or 3/4 occupancy.
---
### Author:  [@uniment](https://discourse.julialang.org/u/uniment)
#### Post date: [March 22, 2023, 12:40am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/4 "2023-03-22T00:40:10Z")
> [@odow](#):
>
> `keys(edge_prob)`, not `edge_prob.keys`
Not confusing at all.
---
### Author:  [@odow](https://discourse.julialang.org/u/odow)
#### Post date: [March 22, 2023, 1:39am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/5 "2023-03-22T01:39:01Z")
> Not confusing at all.
Julia doesn’t have the concept of private fields of a struct. This means accessing `.xxx` fields is usually not intended or safe, unless explicitly documented.
Here’s the documentation for dictionaries: [Collections and Data Structures · The Julia Language](https://docs.julialang.org/en/v1/base/collections/#Dictionaries)
`keys` is documented:
```Julia
julia> d = Dict(1 => 2)
Dict{Int64, Int64} with 1 entry:
1 => 2
help?> keys(d)
keys(a::AbstractDict)
Return an iterator over all keys in a dictionary. collect(keys(a)) returns an array of keys. When
the keys are stored internally in a hash table, as is the case for Dict, the order in which they
are returned may vary. But keys(a) and values(a) both iterate a and return the elements in the
same order.
Examples
≡≡≡≡≡≡≡≡≡≡
julia> D = Dict('a'=>2, 'b'=>3)
Dict{Char, Int64} with 2 entries:
'a' => 2
'b' => 3
julia> collect(keys(D))
2-element Vector{Char}:
'a': ASCII/Unicode U+0061 (category Ll: Letter, lowercase)
'b': ASCII/Unicode U+0062 (category Ll: Letter, lowercase)
```
---
### Author:  [@CameronBieganek](https://discourse.julialang.org/u/CameronBieganek)
#### Post date: [March 22, 2023, 2:00am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/6 "2023-03-22T02:00:32Z")
> [@uniment](#):
>
> Not confusing at all.
Normally I prefer to avoid leading underscores as much as possible, but in this case it might be a good idea for `Base` to rename that field to `_keys` to avoid this gotcha.
---
### Author:  [@CameronBieganek](https://discourse.julialang.org/u/CameronBieganek)
#### Post date: [March 22, 2023, 2:07am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/7 "2023-03-22T02:07:48Z")
I opened an issue with that suggestion:
>
>
> To avoid new users accidentally accessing the internals of a dictionary via \`d.k…eys\` or \`d.vals\` (e.g., if they're coming from an OOP language), it's probably a good idea to change the names of those fields to \`\_keys\` and \`\_vals\`.
>
> Inspired by this question on Discourse:
> https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422
---
### Author:  [@gdalle](https://discourse.julialang.org/u/gdalle)
#### Post date: [March 22, 2023, 6:04am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/8 "2023-03-22T06:04:25Z")
For your specific use case you may also be interested in [GitHub - JuliaGraphs/SimpleWeightedGraphs.jl: Edge-weighted graphs compatible with Graphs.jl](https://github.com/JuliaGraphs/SimpleWeightedGraphs.jl)
---
### Author:  [@gxyau](https://discourse.julialang.org/u/gxyau)
#### Post date: [March 22, 2023, 8:08am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/9 "2023-03-22T08:08:10Z")
Thanks for replying, I got so used to Python and Java at my work that forgot why I liked Julia in the first place lol.
---
### Author:  [@gxyau](https://discourse.julialang.org/u/gxyau)
#### Post date: [March 22, 2023, 8:10am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/10 "2023-03-22T08:10:32Z")
Ahh, thanks for the information. I did looked into SimpleWeightedGraphs, but I needed it to have more structure than just the edge probability so I chose to work from basic principles. Besides, it’s a good practice for myself anyway. (:
---
### Author:  [@gxyau](https://discourse.julialang.org/u/gxyau)
#### Post date: [March 22, 2023, 8:11am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/11 "2023-03-22T08:11:21Z")
I see, that’s why some of the key-value pairs are even repeated.
---
### Author:  [@gdalle](https://discourse.julialang.org/u/gdalle)
#### Post date: [March 22, 2023, 10:26am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/12 "2023-03-22T10:26:49Z")
I agree that it is excellent practice to code it yourself, but in case you need something more reliable, you can always try [GitHub - JuliaGraphs/MetaGraphsNext.jl: Graphs with vertex labels and metadata in Julia](https://github.com/JuliaGraphs/MetaGraphsNext.jl)
---
### Author:  [@gxyau](https://discourse.julialang.org/u/gxyau)
#### Post date: [March 22, 2023, 10:48am UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/13 "2023-03-22T10:48:35Z")
Thank you. 😃 How is the performance and stability of the package right now though, do you know? I was told that it is an option but not very stable but it was a while ago.
---
### Author:  [@gdalle](https://discourse.julialang.org/u/gdalle)
#### Post date: [March 22, 2023, 12:19pm UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/14 "2023-03-22T12:19:21Z")
I have personally cleaned it up a lot lately, and added more tests for robustness. I’m gonna benchmark it against the historical MetaGraphs.jl in the next few days, but I’m fairly confident that its performance is much better
---
### Author:  [@Palli](https://discourse.julialang.org/u/Palli)
#### Post date: [March 22, 2023, 7:09pm UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/15 "2023-03-22T19:09:00Z")
There are also other dictionaries to be aware of, and likely fix, e.g. (SwissDict and RobinDict and) most importantly OrderedDicts (and OrderedSet):
[https://juliacollections.github.io/DataStructures.jl/latest/ordered\_containers/](https://juliacollections.github.io/DataStructures.jl/latest/ordered_containers/)
I suppose you would want PR for such packages too, and at least it’s only a breaking change packages(s), not Julia per se. There’s at least one more dictionary package too, Dictionary.jl.
It might be a good start, to change packages, at least to see how tolerant users are to change, and get them ready.
That said, would there be a way to make `my_dict.keys` work when keys is gone (without a macro?), to get you `keys(my_dict)`. I believe you can overload the period. I believe the former does actually allow you to change the keys, but I would be ok with read-only, likely covers all uses (and it’s a bug anyway as is). Does the later give you keys you can then change?
---
### Author:  [@GunnarFarneback](https://discourse.julialang.org/u/GunnarFarneback)
#### Post date: [March 22, 2023, 7:37pm UTC](https://discourse.julialang.org/t/dictionary-showing-invalid-keys/96422/16 "2023-03-22T19:37:57Z")
> [@Palli](#):
>
> at least it’s only a breaking change packages(s)
It’s not a breaking change to rename internal fields. In this case it’s also pretty difficult to make use of the `keys` field so the probability that some user has mistakenly (and successfully) dipped into the internals is pretty low.
> [@Palli](#):
>
> That said, would there be a way to make `my_dict.keys` work when keys is gone (without a macro?), to get you `keys(my_dict)`.
Yes, you could but why would you want to?