Regarding data frame column type change in DM visual about dm

[TSchiefer]

Thanks, your example would work, if the altered table was used in the dm. Actually, in the function dm_nycflights13(), the flights table is supplied via nycflights13::flights (you can check the function definition by calling dm_nycflights13 without braces in the console), so it will always use the table as it is defined in the {nycflights13} package.

You could make it work here like this:

library(dm)

as_timestamp_without_timezone <- function(x) {
  class(x) <- "timestamp without timezone"
  x
}
nyc_dm <- dm_nycflights13() %>% 
  dm_zoom_to(flights) %>% 
  mutate(dep_time = as_timestamp_without_timezone(dep_time)) %>% 
  dm_update_zoomed()
sapply(nyc_dm$flights, class)
#> $year
#> [1] "integer"
#> 
#> $month
#> [1] "integer"
#> 
#> $day
#> [1] "integer"
#> 
#> $dep_time
#> [1] "timestamp without timezone"
#> 
#> $sched_dep_time
#> [1] "integer"
#> 
#> $dep_delay
#> [1] "numeric"
#> 
#> $arr_time
#> [1] "integer"
#> 
#> $sched_arr_time
#> [1] "integer"
#> 
#> $arr_delay
#> [1] "numeric"
#> 
#> $carrier
#> [1] "character"
#> 
#> $flight
#> [1] "integer"
#> 
#> $tailnum
#> [1] "character"
#> 
#> $origin
#> [1] "character"
#> 
#> $dest
#> [1] "character"
#> 
#> $air_time
#> [1] "numeric"
#> 
#> $distance
#> [1] "numeric"
#> 
#> $hour
#> [1] "numeric"
#> 
#> $minute
#> [1] "numeric"
#> 
#> $time_hour
#> [1] "POSIXct" "POSIXt"

^{Created on 2022-11-29 by the reprex package (v2.0.1)}

[sound118]

Awesome, that works!

[sound118]

Sorry, I still have a puzzle related to this column type issue. Let's take below minimal example:
df1 <- data.frame(
id = c('122','345', '43'),
name = c('john','matt','roger'),
race = c('1','2','1'),
age = c('20','23','34'),
height = c('6.4', '5.7', '4.9'),
stringsAsFactors = FALSE
)

cols <- data.frame(
name = c('id','name', 'race', 'age', 'height'),
type = c('varchar(20)', 'varchar(200)', 'varchar(10)', 'int', 'numeric(3,1)'),
stringsAsFactors = FALSE)

I actually want to convert all column types in DM object into the type in "cols" data frame. I don't want to do this with mutate for each column one by one which is some sort of hard coding. Instead, I am wondering if we could do as following duck-typed script:
dm_obj <- dm(df1)

convert_col_type <- function(df) {
for (i in seq_len(length(colnames(df)))) {
class(df[[colnames(df)[i]]]) <- cols$type[i]
}
df
}

test_dm <- dm_obj %>%
dm_zoom_to(df1) %>%
convert_col_type() %>%
dm_update_zoomed()

Above scrip obviously has some error issue. But if we simply run str(convert_col_type(df1)), it could show us the converted column types. Just wondering how we could incorporate this into "dm" package framework. Thanks a lot in advance!

[sound118]

@TSchiefer hi, any suggestion for above raised issue.?

[TSchiefer]

Sorry for the late answer.
I generally think, not everything needs to be done within a dm object. You could work around doing the following:

suppressPackageStartupMessages({
  library(tibble)
  library(purrr)
  library(dplyr)
  library(dm)
})

df1 <- data.frame(
  id = c('122','345', '43'),
  name = c('john','matt','roger'),
  race = c('1','2','1'),
  age = c('20','23','34'),
  height = c('6.4', '5.7', '4.9'),
  stringsAsFactors = FALSE
)

cols <- data.frame(
  name = c('id','name', 'race', 'age', 'height'),
  type = c('varchar(20)', 'varchar(200)', 'varchar(10)', 'int', 'numeric(3,1)'),
  stringsAsFactors = FALSE) %>% 
  deframe()

# now a named character vector
cols
#>             id           name           race            age         height 
#>  "varchar(20)" "varchar(200)"  "varchar(10)"          "int" "numeric(3,1)"

df2 <- imap(df1, function(col, colname) {
  class(col) <- cols[[colname]]
  col
}) %>% 
  bind_cols()

df2
#> # A tibble: 3 × 5
#>   id         name       race       age   height    
#>   <vrch(20)> <vrc(200)> <vrch(10)> <int> <nmr(3,1)>
#> 1 122        john       1          20    6.4       
#> 2 345        matt       2          23    5.7       
#> 3 43         roger      1          34    4.9
map_chr(df2, class)
#>             id           name           race            age         height 
#>  "varchar(20)" "varchar(200)"  "varchar(10)"          "int" "numeric(3,1)"

dm(df2)
#> ── Metadata ────────────────────────────────────────────────────────────────────
#> Tables: `df2`
#> Columns: 5
#> Primary keys: 0
#> Foreign keys: 0

^{Created on 2022-12-20 by the reprex package (v2.0.1)}