Open johnmyleswhite opened 11 years ago
Ok, thanks. Currently the data is being parsed through an awkward combination of non-general control flow. It works, but only for one value column and I'm not capturing the column name at present.
Brain storm
readtable(s::String, time=true)
Where the first column is parsed as a Calendar object. Should I open an issue?
If we start to do automatic time series conversion, I think we shouldn't hardcore which columns are time series, but either (a) guess from the inputs or (b) require that a list of column names be specified by the user.
read.zoo
in all its kwarg glory
> args(read.zoo)
function (file, format = "", tz = "", FUN = NULL, regular = FALSE,
index.column = 1, drop = TRUE, FUN2 = NULL, split = NULL,
aggregate = FALSE, ..., text)
The format
arg is a string to pass to the Calendar.parse
method. tz
allows one to over-ride the local machine, index.column
defaults to the first one for time, which is good for well over 98% of time series data. I'd pick all three for a readtable
method.
The readable
method would solve this issue as well as make the code for Quandl
a lot neater (and reliable given the trouble with missing values).
Currently, readable
does not support http requests but it appears to be stubbed out to do this in the future.
We still need to finish integrating a generic CSV parser into Base, but it seems that you'd be better off using a general-purpose parser than doing comma-splitting by hand. There's probably a while before the generic parser works well enough for your use, but I thought it would be good to raise the idea now while you're working on your package.