Descending sort by value of a Hash in Ruby

Question

My input hash: h = { "a" => 20, "b" => 30, "c" => 10 }

Ascending sort: h.sort {|a,b| a[1]<=>b[1]} #=> [["c", 10], ["a", 20], ["b", 30]]

But, I need [["b", 30], ["a", 20], ["c", 10]]

How is can we make it work the other way around, what does <=> mean?

Solution 1

You can have it cleaner, clearer and faster, all at once! Like this:

h.sort_by {|k,v| v}.reverse

I benchmarked timings on 3000 iterations of sorting a 1000-element hash with random values, and got these times:

h.sort {|x,y| -(x[1]<=>y[1])} -- 16.7s
h.sort {|x,y| y[1] <=> x[1]} -- 12.3s
h.sort_by {|k,v| -v} -- 5.9s
h.sort_by {|k,v| v}.reverse -- 3.7

Solution 2

h.sort {|a,b| b[1]<=>a[1]}

Solution 3

<=> compares the two operands, returning -1 if the first is lower, 0 if they're equal and 1 if the first is higher. This means that you can just do -(a[1]<=>b[1]) to reverse the order.

Solution 4

Super simple: h.sort_by { |k, v| -v }

View more solutions

Comments

[PJP]

I always prefer seeing it written with the "a" and "b" elements swapped, rather than negating the result. With them swapped I only have to look at their order to see they're backwards to know it's reversed. When the value of <=> is negated I still have to look at the actual comparison to know what is going on. It's a minor point but something I'm aware of because I can feel my brain have to do the second check after doing a "What!?"

[PJP]

Visually this is cleaner but it causes an extra traversal of the collection to reverse it.

[Chuck]

@Greg: I definitely see why you'd prefer it the other way. I'm the opposite: To my scanning eyes, b[1]<=>a[1] looks hella like a[1]<=>b[1] and I feel a need to stop and check, whereas the negation makes it immediately obvious that we're doing a reverse sort.

[PJP]

I understand your point too. Either way of doing it still requires a careful look at the values being compared. Maybe we need a different operator - >=< for reverse order? Nah, that'd be just as bad. It's the entire construct, but I prefer <=> over a more verbose approach where we'd have to call some method names.

[Taryn East]

I've downvoted this comment, not because the answer is wrong, but because you don't explain why this is the correct answer. The questioner even asked specifically what "<=>" means - so (s)he's clearly after some explanation of how this all works. It's a good idea to help out that way :)

[Taryn East]

PS: see the Stack Overflow review policy on meta for more: meta.stackexchange.com/questions/74194/…

[boddhisattva]

@glennmcdonald can you please tell me how can one calculate the timings for each call in ruby ?

[boddhisattva]

I found out how to calculate the time wrt each call. Here's a sample link:- ruby-doc.org/core-1.9.3/Enumerable.html#method-i-sort

[Cruz Nunez]

What about h.sort_by(&:last).reverse? Any times on that?

[danielricecodes]

Which Ruby version?

[Eric Duminil]

@TamzinBlake: I really dont think sort_by mutates the hash. What did you mean exactly?

[Tamzin Blake]

Nice catch! .sort_by! mutates the hash, .sort_by does not.