Output format of reshape functions

[garborg]

@tshort I got familiar with the reshape functions in #644, and had a couple questions about the output format:

1. Placing the id vars to the right of the measure vars -- seemed the opposite of what I'd used elsewhere -- what's the rationale?
2. Trivial: lowercase variable and value -- are they just from before the uppercasing idiom took hold, or is it a reducing clashes / purposely stand out thing?

julia> d = dataset("reshape2", "smiths")
julia> melt(d, [:Subject, :Time], [:Age, :Weight, :Height])
6x4 DataFrame
|-------|----------|-------|--------------|------|
| Row # | variable | value | Subject      | Time |
| 1     | Age      | 33.0  | "John Smith" | 1    |
| 2     | Age      | NA    | "Mary Smith" | 1    |
| 3     | Weight   | 90.0  | "John Smith" | 1    |
| 4     | Weight   | NA    | "Mary Smith" | 1    |
| 5     | Height   | 1.87  | "John Smith" | 1    |
| 6     | Height   | 1.54  | "Mary Smith" | 1    |

[tshort]

Sorry I missed this @garborg. I'm fine with both of your suggested changes.

[bkamins]

We need to decide on this before 1.0 release

[bkamins]

@nalimilan - I would close it. These changes would be breaking and do not improve things significantly enough to justify it I think.

[nalimilan]

Maybe let's just add columns to the end instead of the beginning? That's what the tidyverse (both reshape2 and tidyr) and Pandas do, and it shouldn't break a lot of code anyway since names are more common to use than positions.

[bkamins]

So we switch to: id, variable, value order?

[bkamins]

If yes I will implement this after #2140 is merged (because a lot of documentation will get updated)

[nalimilan]

Yes that sounds like the common order.