Various enhancements to print.data.table

[MichaelChirico]

Current task list:

• Add .Rd file for print.data.table | 045ba8a
• [R-Forge #1957] Optional Arguments for Printing #645 | Ability to turn off row numbers
  ac2180e
• [R-Forge #1957] Optional Arguments for Printing #645 | Ability to turn off smart table wrapping
• [R-Forge #1957] Optional Arguments for Printing #645 | Ability to force-print all entries | a82bfff820d77fbb7ba445e65972e059e036f5150
• [R-Forge #1957] Optional Arguments for Printing #645 | Ability to demarcate by-groupings
• [R-Forge #1957] Optional Arguments for Printing #645 | Demarcation of table border
• [R-Forge #1957] Optional Arguments for Printing #645 | Demarcation of key columns
• print.data.table : expose options('data.table.print.row.names') #1097 | Closes #1097 (progress towards #1523), creates option for printing row names #1570
• FR: Option to exclude the lower register of column names in printing #1482 | Closes #1482, adds options for controlling how print displays column names #2296
• Option for dplyr-like printing [see below]
• [Request] head.data.table print columns fit to screen width #1497 | [REQUEST] Print only first/last few columns #2608 | Print Option: Print columns that fit in single console without wrapping #4074
• 'datatable.prettyprint.char' wrapper #1374
• [R-Forge #5253] printing data.table with no column names #545 Closes #545, with progress towards #1523 -- print handles blank-named tables properly #1574
• [R-Forge #2392] Add length postfix to list column output #605 | trivial (temporary?) fix for confusing list column printing, related,… #2562
• oldtncol (0) < oldncol (5) but tl of class is marked. #2482
• Print Option: Print columns that fit in single console without wrapping #4074
• Option to print the number of columns #6663

---

Some Notes

3 (tabled pending clarification)

As I understand it, this issue is a request to prevent the console output from wrapping around (i.e., to force all columns to appear parallel, regardless of how wide the table is).

If that's the case, this is (AFAICT) impossible, since that's something done by RStudio/R itself. I for one certainly don't know of any way to alter this behavior.

If someone does know of a way to affect this, or if they think I'm mis-interpreting, please pipe up and we can have this taken care of.

7

As I see it there are two options here. One is to treat all key columns the same; the other is to treat secondary, tertiary, etc. keys separately.

Example output:

set.seed(01394)
DT <- data.table(key1 = rep(c("A","B"), each = 4),
                 key2 = rep(c("a","b"), 4),
                 V1 = nrorm(8), key = c("key1","key2"))

# Only demarcate key columns
DT
#    | key1 | | key2 |         V1
#1: |    A | |    a |  0.5994579
#2: |    A | |    a | -1.0898775
#3: |    A | |    b | -0.2285326
#4: |    A | |    b | -1.7858472
#5: |    B | |    a | -0.6269875
#6: |    B | |    a | -0.6633084
#7: |    B | |    b |  1.0367084
#8: |    B | |    b |  0.7364276

# Separately "emboss" keys based on key order
DT
#    | key1 | || key2 ||         V1
#1: |    A | ||    a ||  0.5994579
#2: |    A | ||    a || -1.0898775
#3: |    A | ||    b || -0.2285326
#4: |    A | ||    b || -1.7858472
#5: |    B | ||    a || -0.6269875
#6: |    B | ||    a || -0.6633084
#7: |    B | ||    b ||  1.0367084
#8: |    B | ||    b ||  0.7364276

And of course, add an option for deciding whether to demarcate with | or some other user's-choice character (*, +, etc.)

9 [DONE]

Some feedback from a closed PR that was a first stab at solving this:

From Arun regarding preferred options:

col.names = c("auto", "top", "none")

"auto": current behaviour

"top": only on top, data.frame-like

"none": no column names -- exclude rows in which column names would have been printed.

10 [DONE]

It would be nice to have an option to print a row under the row of column names which gives each column's stored type, as is currently (I understand) the default for the output of dplyr operations.

Example from dplyr:

library(dplyr)
DF <- data.frame(n = numeric(1), c1 = complex(1), i = integer(1),
                 f = factor(1), D = as.Date("2016-02-06"), c2 = character(1),
                 stringsAsFactors = FALSE)
tbl_df(DF)
# Source: local data frame [1 x 6]
#
#       n     c1     i      f          D    c2
#   (dbl) (cmpl) (int) (fctr)     (date) (chr) # <- this row
#1     0   0+0i     0      1 2016-02-06      

Current best alternative is to do sapply(DF, class), but it's nice to have a preview of the data wit this extra information.

11

This seems closely related to 3. Current plan is to implement this as an alternative to 3 since it seems more tangible/doable.

Via @nverno:

Would it be useful for head.data.table to have an option to print only the head of columns that fit the screen width, and summarise the rest? I was imagining something like the printed output from the head of a tbl_df in dplyr. I think it is nice for tables with many columns.

and the guiding example from Arun:

require(data.table)
dt = setDT(lapply(1:100, function(x) 1:3))
dt
dplyr::tbl_dt(dt)

12

Currently covered by @jangorecki's PR #1448; Jan, assuming #1529 is merged first, could you edit the print.data.table man page for your PR?

[arunsrinivasan]

Just brilliant!

[arunsrinivasan]

No idea about 3 and 5 (as to what they mean).
I think a PR for 6 would be nice (seems straightforward from what Jan wrote there). Perhaps ?print.data.table is the time consuming part? Do you think you'd be up for this, @MichaelChirico ?
No idea as to what 7 means either..
8 is another great idea. PR would be great!

[arunsrinivasan]

It'd be really nice if Github would allow assigning tasks to project who aren't necessarily members :-(.

[arunsrinivasan]

There's also #1497

[MichaelChirico]

@arunsrinivasan should I try and PR this one issue at a time? Or in a fell swoop? I've got 8 basically taken care of, just need to add tests.

[arunsrinivasan]

Michael, separate PRs.

[nverno]

Very nice! Sorry to get back to you late on this, but Arun provided a nice example. It is just a nice convenience when interactively looking at tables with lots columns so your console isn't engulfed by a huge data dump when you take a look at the head. Ill close that other one.