Data Reporting Functions • datagoodr

A Research Guide is not a monolith — it is assembled from small, composable reporting functions, one stage at a time. This vignette opens the hood and runs the whole pipeline on a single raw vector: guess its type, measure it, report the numbers as a table, and draw the visual. create_dgf() does exactly this for every column of a dataset at once; here we do it for one, by hand.

library(datagoodr)
v <- dg_demo$revenue   # a raw vector: dollar amounts
head(v)

#> [1]  806.37  310.01  961.78  207.80 1390.73  766.73

Stage 1 — guess the data type

Typing is the first move, and it comes from the companion package formex. Hand it the vector (the column name is a useful hint) and read the $route:

route <- formex::guess_data_type(v, name = "revenue")$route
c(data_type = route$data_type, semantic_type = route$semantic_type,
  depth = route$depth, confidence = round(route$confidence, 2))

#>     data_type semantic_type         depth    confidence 
#>      "number" "measurement"           "3"        "0.98"

A number, recognized from the values. The type is what selects every reporting function that follows.

Stage 2 — measure the column

The measurement functions take a variable name and a data frame and return the profile as JSON — the exact payload datagoodr stores in the DGF’s rg_* cells. The universal quality counts:

df <- data.frame(revenue = v, stringsAsFactors = FALSE)
get_properties("revenue", df)

#> [ 
#>   {  "STAT" :  "Rows"  ,  "VAL" :  "60"  ,  "PER" :  ""  }, 
#>   {  "STAT" :  "Distinct"  ,  "VAL" :  "60"  ,  "PER" :  "(100.0%)"  }
#> ]

and the type-specific statistics (here, the numeric quantile ladder plus mean, skew, and kurtosis):

get_stats("revenue", df, "numeric")

#> [ 
#>   {  "STAT" :  "Minimum"  ,  "VAL" :  "55.91"  }, 
#>   {  "STAT" :  "Q - 05"  ,  "VAL" :  "107.08"  }, 
#>   {  "STAT" :  "Q - 25"  ,  "VAL" :  "207.71"  }, 
#>   {  "STAT" :  "Median"  ,  "VAL" :  "338.8"  }, 
#>   {  "STAT" :  "Mean"  ,  "VAL" :  "534.83"  }, 
#>   {  "STAT" :  "Q - 75"  ,  "VAL" :  "743.11"  }, 
#>   {  "STAT" :  "Q - 95"  ,  "VAL" :  "1,405.33"  }, 
#>   {  "STAT" :  "Maximum"  ,  "VAL" :  "2,100.88"  }, 
#>   {  "STAT" :  "Skew"  ,  "VAL" :  "1.44"  }, 
#>   {  "STAT" :  "Kurtosis"  ,  "VAL" :  "1.58"  }
#> ]

This JSON is the “inner” profile. Nothing about it is tied to how it will be shown — reporting is a separate stage.

Stage 3 — report the numbers as a table

To render, the profile is carried in a DGF and handed to a dg_* reporting function, which parses the JSON and prints a table. Build a one-column DGF and report the stats:

dgf <- create_dgf(df, file = tempfile("dgf"), open = FALSE)

dg_properties(dgf, "revenue")   # the universal counts, as a table

PROPERTIES:

STAT	VAL	PER
Rows	60
Distinct	60	(100.0%)

dg_stats(dgf, "revenue")        # the numeric statistics, as a table

STATS:

STAT	VAL
Minimum	55.91
Median	338.8
Mean	534.83
Maximum	2,100.88
Skew	1.44
Kurtosis	1.58

Stage 4 — create the visual

The graphic is chosen by type too: dg_graphic() dispatches a number to the histogram renderer, reading the same stored profile.

dg_graphic(dgf, "revenue", label = "")

Histogram of the revenue vector

Putting it together

Four stages — guess, measure, tabulate, visualize — each a small function selected by the column’s type. A different vector would route through different functions: a categorical column to paste_treemap, a date to the calendar heat-map, free text to the word cloud. That type-to-function mapping is itself a setting you can inspect and change — see Report Aesthetics and Customization.

create_dgf() runs stages 1–2 for every column and stores the result; the Research Guide runs stages 3–4. Everything in this vignette is what happens inside those two calls.