Add a function for combine/vcat with source names

[johnmyleswhite]

Should we add a function like vcat(df1, df2, df3, source = ["X", "Y", "Z"]) which concatenates DataFrames while also labeling their origins? In this example, the output would be equivalent to doing:

df1_alt = copy(df1)
df1_alt[:source] = "X"
df2_alt = copy(df2)
df2_alt[:source] = "Y"
df3_alt = copy(df3)
df3_alt[:source] = "Z"

vcatted = vcat(df1_alt, df2_alt, df3_alt)

The idea is to preserve the semantics of Base.vcat, but allow the introduction of a custom column (whose name comes from a keyword arg) that denotes the origin of each subset of data.

[HarlanH]

I like the idea.

Could also/instead be vcat(["X" => df1, "Y" => df2, "Z" => df3])
maybe...?

It might also be framable in the split-apply-combine framework as some
variation on just a combine step.

On Sun, Jul 27, 2014 at 11:49 AM, John Myles White <notifications@github.com

wrote:

Should we add a function like vcat(df1, df2, df3, source = ["X", "Y",
"Z"]) which concatenates DataFrames while also labeling their origins? In
this example, the output would be equivalent to doing:

df1_alt = copy(df1)
df1_alt[:source] = "X"
df2_alt = copy(df2)
df2_alt[:source] = "Y"
df3_alt = copy(df3)
df3_alt[:source] = "Z"

vcatted = vcat(df1_alt, df2_alt, df3_alt)

The idea is to preserve the semantics of Base.vcat, but allow the
introduction of a custom column (whose name comes from a keyword arg) that
denotes the origin of each subset of data.

—
Reply to this email directly or view it on GitHub
#659.

[johnmyleswhite]

If you do vcat(["X" => df1, "Y" => df2, "Z" => df3]), how do you figure out what the extra column' name should be?

[HarlanH]

name="source" default argument? I'd mildly prefer that, as separating the
names from the values seems less elegant then putting them next to each
other. Or maybe support both, to preserve the Base.vcat semantics? Or
what about vcat(X = df1, Y = df2, Z = df3)? All optional arguments isn't
great for multiple dispatch though...

On Sun, Jul 27, 2014 at 1:53 PM, John Myles White notifications@github.com
wrote:

If you do vcat(["X" => df1, "Y" => df2, "Z" => df3]), how do you figure
out what the extra column' name should be?

—
Reply to this email directly or view it on GitHub
#659 (comment)
.

[jwmerrill]

I was just looking at the tidy data paper, and it sounds like this is similar to ldply from plyr.

[quinnj]

Anyone still interested in this? Could be useful if someone wants to take a stab at it. @cjprybol ?

[bkamins]

@oxinabox seems to also want it in join 😄.