Created CJ.dt - for cross joining two data.tables

[stephlocke]

Done as separate code on the assumption that it can later be deprecated when better code is integrated into CJ to handle two data.tables

Was initially going to be a helper function in optiRum but getting it into data.table is much more preferable. I've not included any unit tests as the difference is significant but please see these testthat unit tests for it I've written for the function

[eantonya]

The sorting of your function does not match that of CJ itself.

On Tue, Sep 16, 2014 at 2:50 PM, stephlocke notifications@github.com
wrote:

Done as separate code on the assumption that it can later be deprecated
when better code is integrated into CJ to handle two data.tables

Was initially going to be a helper function in optiRum but getting it into
data.table is much more preferable. I've not included any unit tests as the
difference is significant but please see these testthat unit tests for it
https://github.com/stephlocke/optiRum/blob/master/tests/testthat/test-CJ.dt.R

I've written for the function

You can merge this Pull Request by running

git pull https://github.com/stephlocke/data.table master

Or view, comment on, or merge it at:

#814
Commit Summary

• Create CJ.dt

File Changes

• A R/CJ.dt
  https://github.com/Rdatatable/data.table/pull/814/files#diff-0 (26)

Patch Links:

• https://github.com/Rdatatable/data.table/pull/814.patch
• https://github.com/Rdatatable/data.table/pull/814.diff

—
Reply to this email directly or view it on GitHub
#814.

[eantonya]

Here's a suggestion for how to do it using CJ, that can take an arbitrary
number of data.tables or vectors:

CJ.dt = function(...) {
rows = do.call(CJ, lapply(list(...), function(x) if(is.data.frame(x)) seq_len(nrow(x)) else seq_along(x)));
do.call(data.table, Map(function(x, y) x[y], list(...), rows))
}

(the subset part above can almost certainly be optimized)

On Tue, Sep 16, 2014 at 3:09 PM, Eduard Antonyan eduard.antonyan@gmail.com
wrote:

The sorting of your function does not match that of CJ itself.

On Tue, Sep 16, 2014 at 2:50 PM, stephlocke notifications@github.com
wrote:

Done as separate code on the assumption that it can later be deprecated
when better code is integrated into CJ to handle two data.tables

Was initially going to be a helper function in optiRum but getting it
into data.table is much more preferable. I've not included any unit tests
as the difference is significant but please see these testthat unit
tests for it
https://github.com/stephlocke/optiRum/blob/master/tests/testthat/test-CJ.dt.R

I've written for the function

You can merge this Pull Request by running

git pull https://github.com/stephlocke/data.table master

Or view, comment on, or merge it at:

#814
Commit Summary

• Create CJ.dt

File Changes

• A R/CJ.dt
  https://github.com/Rdatatable/data.table/pull/814/files#diff-0 (26)

Patch Links:

• https://github.com/Rdatatable/data.table/pull/814.patch
• https://github.com/Rdatatable/data.table/pull/814.diff

—
Reply to this email directly or view it on GitHub
#814.

[stephlocke]

You're right it doesn't, it does not do any sorting and leaves data in the order of that the data.tables are already in.

Totally appreciate that the behaviour of CJ is to apply sorted=TRUE for backwards compatibility - it could do some sort of default sort, or the sorted argument could be made optional but with a default if/when CJ.dt was integrated?

[stephlocke]

Just saw your suggestion - yes, making a more generic function is a much better idea 👍

[skanskan]

R documentation speaks about CJ.dt
https://www.rdocumentation.org/packages/optiRum/versions/0.40.1/topics/CJ.dt

But when I try o use it I get the error

Error in CJ.dt(z1, z2) : could not find function "CJ.dt"

[MichaelChirico]

@skanskan I guess this is filed on the wrong issue tracker?

https://github.com/lockedata/optiRum/issues

I see CJ.dt in the NAMESPACE so maybe install from dev?

[stephlocke]

Yes - I implemented this in optiRum initially and look forward to it being deprecated one day :)

[skanskan]

Anyway I'm getting the result I want with
myDT[,myvector, by=c("all columns")]
It works well if my vector is constant.

[MichaelChirico]

Hmm I guess this is #1717