setclass() added to data.table.r

[rsaporta]

Motivated by the question raised in #839, it might be useful to have a quick wrapper for

  setattr(x, "class", c(newclass, class(x))

[rsaporta]

If others agree it's useful, I'll go ahead and write up news/doc

[gsee]

I love it, but I think append should be FALSE by default.

[rsaporta]

@gsee, you are probably right.

In my Rprofile, I have this function named as appendClass_( ) But I personally use the convention of ending function names with _ when modifying by reference. In this package, the convetion is to name the function set*.

Would it make sense to use an alternate name with append=TRUE by default?
Or perhaps just leave it as setclass and change default behavior

[gsee]

agreed. With the current implementation, if I call setclass(DT, "data.table") it will give a warning and then call setDT(DT). setDT has this code: if (is.data.table(x)) return(invisible(x)). Therefore, if you set the class of an object to "data.table" (instead of c("data.table", "data.frame")), then calling setclass(DT, "data.table") or setDT(DT) will return an object that is not a data.frame. I think setDT should actually return a c("data.table", "data.frame") even if it is given a "data.table"

[rsaporta]

Ah, I think I see where the confusion is:

then calling setclass(DT, "data.table") or setDT(DT) will return an object that is not a data.frame.

I do not think that is the case.
Anytime is.data.table(x) == TRUE then is.data.frame(x) must be TRUE -- unless of course the user specifically removed the data.frame class.

so in the case of if (is.data.table(x)) return(invisible(x)) , if we look at is(x) I am sure we will see something like c(data.table, data.frame, list, etc)

[rsaporta]

Separately, perhaps the warning in setclass(x, "data.table") is unnecessary. We can simply call setDT() quietly

[gsee]

> DF <- data.frame(A=1:5)
> setattr(DF, "class", "data.table")
> class(DF)
[1] "data.table"
> is(DF)
[1] "data.table" "data.frame" "list"       "oldClass"   "vector"  

But...

> is.data.table(DF)
[1] TRUE
> is.data.frame(DF)
[1] FALSE

and

> setDT(DF)
> class(DF)
[1] "data.table"
> is.data.frame(DF)
[1] FALSE

[gsee]

So I think it'd be important to guard against doing this accidentally

I think I misunderstood. You're saying the user should guard against making the accident, not that setDT() and/or setclass() should try to fix it?

[rsaporta]

no, @gsee, you had understood correctly. I had misunderstood. Specifically, I didn't realize that setDT(DF); is.data.frame(DF) returned FALSE.

Personally, I think it should return TRUE and that setDT() might need to be modified, unless there may be some reason not to do this, of which I am unaware.

@mattdowle and/or @arunsrinivasan , thoughts regarding output of setDT()

[arunsrinivasan]

@rsaporta right, setDT could fix the class for the case you and Garrett mention.

But I'm not sure why we need setclass. setnames is specifically for data.frames and data.tables only. And to convert a data.frame to data.table, we've setDT now, which is easier than doing setclass(DF, c("data.table", "data.frame")).

[rsaporta]

@Arun, I use a function like setclass often. Usually, for inheriting from data.table. (Eg, I may use something like setclass(DT.qres, "query.results") and then print(DT.qres) deploys the appropriate print function.

It's just a handy wrapper to save a few keystrokes.

[arunsrinivasan]

So setclass works only with data.tables, and would append classes provided as input before the already existing classes by default, and errors when input is not a data.table pointing to setattr?

[jangorecki]

I've run into an error produced (most probably) by altered class of data.table.
It occurs when rendering Rmd file, the chunk in Rmd contains the following:

ticker <- market.api.process("kraken",c("BTC","EUR"),"ticker") # Rbitcoin 0.9.3 function returns DT with additional class
pander::pandoc.table(ticker)

and it results following R Markdown error:

Error in dimnames.data.table(x) : data.table inherits from data.frame (from v1.5) but this data.table does not. Has it been created manually (e.g. by using structure' rather than 'data.table') or saved to disk using a prior version of data.table? The correct class is c('data.table','data.frame'). Calls: <Anonymous> ... has.rownames -> rownames -> dimnames -> dimnames.data.table Execution halted

and the class of ticker:

class(ticker)
# [1] "btc.ticker" "data.table" "data.frame"

Using the following solve the problem, so it must be related to additional class in DT:

pander::pandoc.table(setDT(setDF(ticker)))

also related to already closed #839

[arunsrinivasan]

Closing this, as it's stale at this point. Have added an issue for setclass() function, #1412.