For now that’s the only public interface. It’s technically possible to use any 
seekable IO stream, but that requires knowing implementation details at the 
moment.

 — John

On Jan 15, 2014, at 5:41 PM, James Fairbanks <[email protected]> wrote:

> The documentation shows a constructor DataStream(). Is the only way to get a 
> DataStream to read it from a file using readstream?
> 
> James
> 
> On Thursday, January 9, 2014 11:22:18 AM UTC-5, John Myles White wrote:
> I just restored the DataStream functionality. Still need to revise its 
> documentation, but (assuming you’re on master for both DataArrays and 
> DataFrames) you can try: 
> 
> using DataFrames 
> ds = readstream(“data.csv”, nrows = 100) 
> 
> for minibatch in ds 
>         show(minibatch, true) 
> end 
> 
>  — John 
> 
> On Jan 9, 2014, at 6:59 AM, Iain Dunning <[email protected]> wrote: 
> 
> > So we have readlines, which loads the file all at once into an array of 
> > strings, and eachline, which allows us to iterate through a file line by 
> > line. 
> > 
> > We have readcsv/readdlm, but don't have an iterating version (as far as I 
> > know). This is a bit of a problem because it means I have to manually parse 
> > CSV files that are too big to read in all at once, which is error prone 
> > (e.g. quoted delimiter handling). Is there demand for this, and what would 
> > it be called? 
> > 
> > Cheers, 
> > Iain 
> 

Reply via email to