Allow a default value?

[shashi]

Missing indices now return a KeyError. Let's track the arguments for and against allowing a default value (such as 0) in NDSparse, so that it can behave something like an Array.

Pros:

• More consistency with arrays

Cons:

• Indices then are essentially unbounded (is this a problem in practice?)

Unclear:

• What should join operations do? innerjoin and outerjoin become equivalent since keys can be found in either table.

[aaowens]

I know this is an old issue but I just had need for this feature in my project. I'm looking up a key which may not exist. If it doesn't exist, I would like to obtain missing.

I tried to just implement this manually, but it doesn't seem like haskey is implemented for NDSparse. Was this intentional? I could submit a PR for it.

[joshday]

We should have a haskey method. A PR would be great!

I lean towards get and get! methods over default values. Those should be straightforward to implement as well if you're up for it. Otherwise I'll add them when I can.

[piever]

haskey is nice but I agree that get and get! methods sounds like the ideal solution (as well as get with a function argument to be run only if the value is not found, which we could even use to implement getindex, to avoid code duplication).