patterns in .SDcols

[eantonya]

I thought I've seen this FR before, but couldn't find it.

Would be nice if we could specify column names using regex expressions in .SDcols. Currently one has to do something like .SDcols = grep("mypattern", names(myDT)), which you can't chain on, and is pretty fragile.

Perhaps the patterns function from melt can be reused here, making the syntax .SDcols = patterns("mypattern").

[ksavin]

I'd like to add, that patterns would be super useful in j as well.

It is often needed to select columns with grep and the only way is to refer it via names(), e.g.
veryLongDataTableName[, grep('lag', names(veryLongDataTableName), with = FALSE]

or to remove multiple columns, e.g.
dt[, (grep('lag', names(dt)) := NULL]

or to make new column names:
dt[, paste0(grep('^test', names(dt), value = TRUE), '_sqrt') := lapply(.SD, sqrt), .SDcols = grep('^test', names(dt), value = TRUE)]

These would be much shorter with patterns available in j as well:
veryLongDataTableName[, patterns('lag'), with = FALSE]
dt[, patterns('lag') := NULL]
dt[, paste0(patterns('^test'), '_sqrt') := lapply(.SD, sqrt), .SDcols = patterns('^test')]

Alternatively, it would be handy to have a special symbol for column names, selected in .SDcols, e.g. .NM

[MichaelChirico]

names(.SD) should suffice...

On Oct 26, 2016 12:37 PM, "ksavin" notifications@github.com wrote:

I'd like to add, that patterns would be super useful in j as well.

It is often needed to select columns with grep and the only way is to
refer it via names(), e.g.
veryLongDataTableName[, grep('lag', names(veryLongDataTableName), with =
FALSE]

or to remove multiple columns, e.g.
dt[, (grep('lag', names(dt)) := NULL]

or to make new column names:
dt[, paste0(grep('^test', names(dt), value = TRUE), '_sqrt') :=
lapply(.SD, sqrt), .SDcols = grep('^test', names(dt), value = TRUE)]

These would be much shorter with patterns available in j as well:
veryLongDataTableName[, patterns('lag'), with = FALSE]
dt[, patterns('lag') := NULL]
dt[, paste0(patterns('^test'), '_sqrt') := lapply(.SD, sqrt), .SDcols =
patterns('^test')]

Alternatively, it would be handy to have a special symbol for column
names, selected in .SDcols, e.g. .NM

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#1878 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AHQQdapOU-mUdVF-4ADOsKkuD4cSC3Lvks5q34G7gaJpZM4KXfit
.

[ksavin]

Forgot I am in fact using names(.SD) for these cases :)
Still, way cleaner and shorter with patterns.

[MichaelChirico]

I do regularly things like grep(pattern, names(.SD)) (and maybe add value = TRUE)... Maybe I'm just used to this setup.

[mbacou]

Another one I tend to use is .SDcols=names(.SD) %like% "mypattern", a little ugly. Upvoting this FR as well.

[hannes101]

Just as a reference to a SO question, please update it there also if it got implemented :-)
https://stackoverflow.com/questions/42999949/select-data-table-columns-with-grep-like-partial-matching

[franknarf1]

SO q to update: https://stackoverflow.com/questions/47535845/perform-operations-on-data-table-columns-based-on-regex?noredirect=1

A to update: https://stackoverflow.com/a/51331981/

[HughParsonage]

I wrote select_grep in package hutils before realizing this was an outstanding issue:

library(hutils)
library(data.table)
dt <- data.table(x1 = 1, x2 = 2, y = 0)
select_grep(dt, "x")
#>    x1 x2
#> 1:  1  2
    select_grep(dt, "x", .and = "y")
#>    x1 x2 y
#> 1:  1  2 0
    select_grep(dt, "x", .and = "y", .but.not = "x2")
#>    x1 y
#> 1:  1 0

Created on 2018-05-19 by the reprex package (v0.2.0).