Use tuples in msgpack

[mrocklin]

This makes msgpack decode arrays as tuples rather than lists, allowing them to be used within containers like set and dict that expect hashable elements.

cc @mariusvniekerk

I took a quick look at this but then got side tracked by other things. I thought I'd push it up here in case you were interested in carrying on.

[rbubley]

It seems that the unit tests in msgpack-python use a different approach to the issue of tuples.

See, e.g. lbolla/msgpack-python@c8042ea

[mariusvniekerk]

I like the approach in that and we can implement set, tuple, frozenset which is probably good enough for now? Implemented it in #2003

[mrocklin]

Do we need to move sets with msgpack? Do we need to move anything other than tuples?

…

On Wed, May 23, 2018 at 9:39 PM, Marius van Niekerk < ***@***.***> wrote: I like the approach in that and we can implement set, tuple, frozenset which is probably good enough for now? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2000 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AASszJcSxQsQuvlPzOTdDAcCwUD0XE7_ks5t1g9IgaJpZM4UKY97> .

[mariusvniekerk]

I've fixed some of the failing tests here in a branch of mine. Whats the easiest way to add those commits

[mrocklin]

Because I'm somewhat occupied this week the best solution might be to submit a PR of your own to the master branch, possibly including my commits or not, as you like. You could also submit a PR to my PR's branch.

…

On Tue, May 29, 2018 at 12:19 PM, Marius van Niekerk < ***@***.***> wrote: I've fixed some of the failing tests here in a branch of mine. Whats the easiest way to add those commits — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2000 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AASszGuMRd8u65o7-HHXZ1rC_OftLP7mks5t3XUAgaJpZM4UKY97> .

[mrocklin]

OK, hypothetically this passes tests. There are some things that concern me though:

1. We are inconsistent about our use of tuples and lists. We use tuples everywhere except if the tuple contained a Serilized object, in that case we use a list
2. We actually have to do this, because we only traverse through lists to find serialized objects.

I would not be surprised if this came back to bite us later.

[mrocklin]

Well, lets see what happens. Merging.