ksformat Usage Examples

In ksformat: 'SAS'-Style 'PROC FORMAT' for R

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
library(ksformat)

The ksformat package provides SAS PROC FORMAT-like functionality for R. This vignette walks through the most common use cases.

Example 1: Basic Discrete Formatting

Create a format for gender codes (auto-stored in library as "sex"):

fnew(
  "M" = "Male",
  "F" = "Female",
  .missing = "Unknown",
  .other = "Other Gender",
  name = "sex"
)

gender_codes <- c("M", "F", "M", NA, "X", "F")
formatted_genders <- fput(gender_codes, "sex")

data.frame(
  code = gender_codes,
  label = formatted_genders
)

fprint("sex")

Example 2: Numeric Range Formatting

Define formats in SAS-like text (auto-registered):

fparse(text = '
VALUE age (numeric)
  [0, 18)     = "Child"
  [18, 65)    = "Adult"
  [65, HIGH]  = "Senior"
  .missing    = "Age Unknown"
;
')

ages <- c(5, 15.3, 17.9, 18, 45, 64.99, 65, 85, NA)
age_groups <- fputn(ages, "age")

data.frame(
  age = ages,
  group = age_groups
)

Example 3: Decimal Ranges (BMI Categories)

fparse(text = '
VALUE bmi (numeric)
  [0, 18.5)    = "Underweight"
  [18.5, 25)   = "Normal"
  [25, 30)     = "Overweight"
  [30, HIGH]   = "Obese"
  .missing     = "No data"
;
')

bmi_values <- c(16.2, 18.5, 22.7, 25, 29.9, 35.1, NA)
bmi_labels <- fputn(bmi_values, "bmi")

data.frame(
  bmi = bmi_values,
  category = bmi_labels
)

Example 4: Exclusive/Inclusive Bounds

fparse(text = '
VALUE score (numeric)
  (0, 50]    = "Low"
  (50, 100]  = "High"
  .other     = "Out of range"
;
')

scores <- c(0, 1, 50, 51, 100, 101)
score_labels <- fputn(scores, "score")

data.frame(
  score = scores,
  label = score_labels
)

Example 5: Reverse Formatting with Invalue

Invalues convert labels back to values. The default target_type is "numeric":

finput(
  "Male" = 1,
  "Female" = 2,
  name = "sex_inv"
)

labels <- c("Male", "Female", "Male", "Unknown", "Female")
codes <- finputn(labels, "sex_inv")

data.frame(
  label = labels,
  code = codes
)

Example 6: Bidirectional Formatting

fnew_bid() creates both a format and an invalue at once:

status_bi <- fnew_bid(
  "A" = "Active",
  "I" = "Inactive",
  "P" = "Pending",
  name = "status"
)

# Forward: code -> label
status_codes <- c("A", "I", "P", "A")
status_labels <- fputc(status_codes, "status")
data.frame(code = status_codes, label = status_labels)

# Reverse: label -> code
test_labels <- c("Active", "Pending", "Inactive")
test_codes <- finputc(test_labels, "status_inv")
data.frame(label = test_labels, code = test_codes)

Example 7: Parse Multiple Formats from Text

fparse(text = '
// Study format definitions

VALUE race (character)
  "W" = "White"
  "B" = "Black"
  "A" = "Asian"
  .missing = "Unknown"
;

INVALUE race_inv
  "White" = 1
  "Black" = 2
  "Asian" = 3
;
')

flist()   # character vector of names
fprint()

Example 8: Export Formats Back to Text

bmi_fmt <- format_get("bmi")
cat(fexport(bmi = bmi_fmt))

Example 9: SAS-like PUT/INPUT Functions

# fputn — apply numeric format by name
fputn(c(5, 30, 70), "age")

# fputc — apply character format by name
fputc(c("M", "F"), "sex")

# finputn — apply numeric invalue by name
finputn(c("White", "Black"), "race_inv")

Example 10: Data Frame Formatting

df <- data.frame(
  id = 1:6,
  sex = c("M", "F", "M", "F", NA, "X"),
  age = c(15, 25, 45, 70, 35, NA),
  stringsAsFactors = FALSE
)

sex_f <- format_get("sex")
age_f <- format_get("age")

df_formatted <- fput_df(
  df,
  sex = sex_f,
  age = age_f,
  suffix = "_label"
)

df_formatted

Example 11: Missing Value Handling

# With .missing label
fput(c("M", "F", NA), "sex")

# With keep_na = TRUE
fput(c("M", "F", NA), sex_f, keep_na = TRUE)

# is_missing() checks
is_missing(NA)
is_missing(NaN)
is_missing("")   # TRUE — empty strings are treated as missing

Example 12: Date/Time Formats (SAS-style)

SAS Date Formats

SAS date format names are auto-resolved — no pre-creation needed:

today <- Sys.Date()

data.frame(
  format = c("DATE9.", "MMDDYY10.", "DDMMYY10.", "YYMMDD10.",
             "MONYY7.", "WORDDATE.", "YEAR4.", "QTR."),
  result = c(
    fputn(today, "DATE9."),
    fputn(today, "MMDDYY10."),
    fputn(today, "DDMMYY10."),
    fputn(today, "YYMMDD10."),
    fputn(today, "MONYY7."),
    fputn(today, "WORDDATE."),
    fputn(today, "YEAR4."),
    fputn(today, "QTR.")
  )
)

# Multiple dates
dates <- as.Date(c("2020-01-15", "2020-06-30", "2020-12-25"))
fputn(dates, "DATE9.")

R Numeric Dates (Days Since 1970-01-01)

r_days <- as.numeric(as.Date("2025-01-01"))
r_days
fputn(r_days, "DATE9.")
fputn(r_days, "MMDDYY10.")

Time Formats

Time is represented as seconds since midnight:

seconds <- c(0, 3600, 45000, 86399)

data.frame(
  seconds = seconds,
  TIME8 = fputn(seconds, "TIME8."),
  TIME5 = fputn(seconds, "TIME5."),
  HHMM = fputn(seconds, "HHMM.")
)

Datetime Formats

now <- Sys.time()

data.frame(
  format = c("DATETIME20.", "DATETIME13.", "DTDATE.", "DTYYMMDD."),
  result = c(
    fputn(now, "DATETIME20."),
    fputn(now, "DATETIME13."),
    fputn(now, "DTDATE."),
    fputn(now, "DTYYMMDD.")
  )
)

# From numeric R-epoch seconds
r_secs <- as.numeric(as.POSIXct("2025-06-15 14:30:00", tz = "UTC"))
fputn(r_secs, "DATETIME20.")

Custom Date Formats with fnew_date()

# SAS-named format
fnew_date("DATE9.", name = "bday_fmt")
birthdays <- as.Date(c("1990-03-25", "1985-11-03", "2000-07-14"))
fput(birthdays, "bday_fmt")

# Custom strftime pattern (e.g. DD.MM.YYYY)
fnew_date("%d.%m.%Y", name = "ru_date", type = "date")
fput(birthdays, "ru_date")

# Custom pattern with missing label
fnew_date("MMDDYY10.", name = "us_date", .missing = "NO DATE")
mixed <- c(as.Date("2025-01-01"), NA, as.Date("2025-12-31"))
fput(mixed, "us_date")

fprint("bday_fmt")

Date Formats in Data Frames

patients <- data.frame(
  id = 1:4,
  visit_date = as.Date(c("2025-01-10", "2025-02-15", "2025-03-20", NA)),
  stringsAsFactors = FALSE
)

visit_fmt <- fnew_date("DATE9.", name = "visit_fmt", .missing = "NOT RECORDED")
fput_df(patients, visit_date = visit_fmt)

Parse Date Formats from Text

fparse(text = '
VALUE enrldt (date)
  pattern = "DATE9."
  .missing = "Not Enrolled"
;

VALUE visit_time (time)
  pattern = "TIME8."
;

VALUE stamp (datetime)
  pattern = "DATETIME20."
;
')

fput(as.Date("2025-03-01"), "enrldt")
fput(36000, "visit_time")
fput(as.POSIXct("2025-03-01 10:00:00", tz = "UTC"), "stamp")

# Export back to text
enrl_obj <- format_get("enrldt")
cat(fexport(enrldt = enrl_obj))

fclear()

Example 13: Multilabel Formats

Overlapping Age Categories

With multilabel formats, a single value can match multiple labels:

fnew(
  "0,5,TRUE,TRUE"    = "Infant",
  "6,11,TRUE,TRUE"   = "Child",
  "12,17,TRUE,TRUE"  = "Adolescent",
  "0,17,TRUE,TRUE"   = "Pediatric",
  "18,64,TRUE,TRUE"  = "Adult",
  "65,Inf,TRUE,TRUE" = "Elderly",
  "18,Inf,TRUE,TRUE" = "Non-Pediatric",
  name = "age_categories",
  type = "numeric",
  multilabel = TRUE
)

ages <- c(3, 14, 25, 70)

# fput returns first match only
fput(ages, "age_categories")

# fput_all returns ALL matching labels
all_labels <- fput_all(ages, "age_categories")
for (i in seq_along(ages)) {
  cat("Age", ages[i], "->", paste(all_labels[[i]], collapse = ", "), "\n")
}

Multilabel with Missing Values

fnew(
  "0,100,TRUE,TRUE"   = "Valid Score",
  "0,49,TRUE,TRUE"    = "Below Average",
  "50,100,TRUE,TRUE"  = "Above Average",
  "90,100,TRUE,TRUE"  = "Excellent",
  .missing = "No Score",
  .other = "Out of Range",
  name = "score_ml",
  type = "numeric",
  multilabel = TRUE
)

scores <- c(95, 45, NA, 150)
ml_result <- fput_all(scores, "score_ml")

for (i in seq_along(scores)) {
  cat("Score", ifelse(is.na(scores[i]), "NA", scores[i]),
      "->", paste(ml_result[[i]], collapse = ", "), "\n")
}

Parse Multilabel from Text

fparse(text = '
VALUE risk (numeric, multilabel)
  [0, 3]   = "Low Risk"
  [0, 7]   = "Monitored"
  (3, 7]   = "Medium Risk"
  (7, 10]  = "High Risk"
;
')

risk_scores <- c(2, 5, 9)
risk_labels <- fput_all(risk_scores, "risk")
for (i in seq_along(risk_scores)) {
  cat("Score", risk_scores[i], "->",
      paste(risk_labels[[i]], collapse = " | "), "\n")
}

Multilabel Export

risk_obj <- format_get("risk")
cat(fexport(risk = risk_obj))

fprint("risk")

Practical Example: Adverse Event Severity Grading

fnew(
  "1,1,TRUE,TRUE" = "Mild",
  "2,2,TRUE,TRUE" = "Moderate",
  "3,3,TRUE,TRUE" = "Severe",
  "4,4,TRUE,TRUE" = "Life-threatening",
  "5,5,TRUE,TRUE" = "Fatal",
  "3,5,TRUE,TRUE" = "Serious",
  "1,2,TRUE,TRUE" = "Non-serious",
  name = "ae_grade",
  type = "numeric",
  multilabel = TRUE
)

grades <- c(1, 2, 3, 4, 5)
ae_labels <- fput_all(grades, "ae_grade")
for (i in seq_along(grades)) {
  cat("Grade", grades[i], ":",
      paste(ae_labels[[i]], collapse = " + "), "\n")
}

fclear()

Example 14: Case-Insensitive Matching

sex_nc <- fnew(
  "M" = "Male",
  "F" = "Female",
  .missing = "Unknown",
  name = "sex_nc",
  type = "character",
  ignore_case = TRUE
)

input <- c("m", "F", "M", "f", NA)
fput(input, sex_nc)

# Note the [nocase] flag
fprint("sex_nc")

# Also works with fputc
fputc("m", "sex_nc")

fclear()

Example 15: Expression Labels in Formats

Expression labels contain .x1, .x2, etc., which reference extra arguments passed to fput(). This lets you compute labels dynamically.

Simple sprintf Expression

stat_fmt <- fnew(
  "n"   = "sprintf('%s', .x1)",
  "pct" = "sprintf('%.1f%%', .x1 * 100)",
  name = "stat",
  type = "character"
)

types  <- c("n",  "pct",  "n",   "pct")
values <- c(42,   0.053,  100,   0.255)

fput(types, stat_fmt, values)

Two Extra Arguments (.x1, .x2)

ratio_fmt <- fnew(
  "ratio" = "sprintf('%s/%s', .x1, .x2)",
  name = "ratio",
  type = "character"
)

fput("ratio", ratio_fmt, 3, 10)
fput(c("ratio", "ratio"), ratio_fmt, c(3, 7), c(10, 20))

ifelse Expression

sign_fmt <- fnew(
  "val" = "ifelse(.x1 > 0, paste0('+', .x1), as.character(.x1))",
  name = "sign",
  type = "character"
)

nums <- c(5, 0, -3)
fput(rep("val", 3), sign_fmt, nums)

Mixed Static and Expression Labels

mixed_fmt <- fnew(
  "header" = "HEADER",
  "n"      = "sprintf('N=%s', .x1)",
  "pct"    = "sprintf('%.1f%%', .x1 * 100)",
  name = "mixed",
  type = "character"
)

keys <- c("header", "n", "pct", "header", "n")
vals <- c(0,        42,  0.15,  0,        100)
fput(keys, mixed_fmt, vals)

Expression in .other Fallback

known_fmt <- fnew(
  "ok" = "OK",
  .other = "sprintf('Error(%s)', .x1)",
  name = "err_fmt",
  type = "character"
)

codes   <- c("ok", "E01", "ok", "E99")
details <- c("",   "timeout", "", "overflow")
fput(codes, known_fmt, details)

Scalar Recycling

label_fmt <- fnew(
  "val" = "sprintf('%s (N=%s)', .x1, .x2)",
  name = "recycle",
  type = "character"
)

fput(c("val", "val"), label_fmt, c(42, 55), 100)

Statistical Table Format with Computed Labels

A realistic clinical-trial example: e() marks labels as expressions evaluated at apply-time, .x1 references the extra argument, and multiline dplyr::case_when shows complex conditional formatting.

# Population counts used as denominators
n.trt <- data.frame(pop = c("fas","pps","saf"), ntot = c(34, 30, 36))
get_n <- function(pop) {
  n.trt$ntot[n.trt$pop == pop]
}

fnew(
  "n_fas" = e("get_n('fas')"),
  "n_pps" = e("get_n('pps')"),
  "n_saf" = e("get_n('saf')"),
  "n"   = "sprintf('%d', .x1)",
  "n_pct_fas" = "sprintf('%d (%5.1f%%)', .x1, .x1 * 100 / get_n('fas'))",
  "n_pct_pps" = "sprintf('%d (%5.1f%%)', .x1, .x1 * 100 / get_n('pps'))",
  "n_pct_saf" = "sprintf('%d (%5.1f%%)', .x1, .x1 * 100 / get_n('saf'))",
  "pct" = "dplyr::case_when(
               .x1>0 & .x1<0.1 ~ sprintf('%5s', ' <0.1%'),
               .x1>=0.1 | .x1==0 ~ sprintf(paste0('%5.', 1 ,'f%%'), .x1)
           )",
  "pval" = "dplyr::case_when(
                .x1>=0 & .x1<0.001 ~ sprintf('%s', '<0.001'),
                .x1>=0.001 & .x1<=0.999 ~ sprintf(paste0('%.', 3 ,'f'), .x1),
                .x1>0.999 ~ sprintf('%s', '>0.999'), .default = '--'
           )",
  name = "stat",
  type = "character"
)

The same format can be created via fparse(). Note that multiline expressions must be collapsed to single lines in the text block, and (eval) marks evaluated labels:

fmt <- '
  VALUE stat_01 (character)
     "n_fas" = "get_n(\'fas\')" (eval)
     "n_pps" = "get_n(\'pps\')" (eval)
     "n_saf" = "get_n(\'saf\')" (eval)
     "n"     = "sprintf(\'%d\', .x1)"
     "pct"   = "dplyr::case_when(.x1>0 & .x1<0.1 ~ sprintf(\'%5s\', \' <0.1%\'), .x1>=0.1 | .x1==0 ~ sprintf(paste0(\'%5.\', 1 ,\'f%%\'), .x1))"
     "n_pct_fas" = "sprintf(\'%d (%5.1f%%)\', .x1, .x1 * 100 / get_n(\'fas\'))"
     "n_pct_pps" = "sprintf(\'%d (%5.1f%%)\', .x1, .x1 * 100 / get_n(\'pps\'))"
     "n_pct_saf" = "sprintf(\'%d (%5.1f%%)\', .x1, .x1 * 100 / get_n(\'saf\'))"
     "pval"  = "dplyr::case_when(.x1>=0 & .x1<0.001 ~ sprintf(\'%s\', \'<0.001\'), .x1>=0.001 & .x1<=0.999 ~ sprintf(paste0(\'%.\', 3 ,\'f\'), .x1), .x1>0.999 ~ sprintf(\'%s\', \'>0.999\'), .default = \'--\')"
;'
fparse(fmt)

Both stat (via fnew) and stat_01 (via fparse) produce identical results:

df <- data.frame(
  types = c("n_fas", "n_pps", "n_saf", "n", "pct", "pct", "n", "pval", "pval",
            "n_pct_fas", "n_pct_pps", "n_pct_saf"),
  values = c(NA, NA, NA, 42, 0.053, 0.0008, 100, 0.255, 0.0003, 22, 22, 22)
)

df$fmt    <- fput(df$types, "stat",    df$values)
df$fmt_01 <- fput(df$types, "stat_01", df$values)
print(df)

fclear()

Example 16: Vectorized Format Names (SAS PUTC-style)

Each element can use a different format, determined by a vector of format names:

# Dispatch format: maps type code to format name
fnew("1" = "groupx", "2" = "groupy", "3" = "groupz",
     name = "typefmt", type = "numeric")

# Per-group character formats
fnew("positive" = "agree",  "negative" = "disagree", "neutral" = "notsure",
     name = "groupx", type = "character")
fnew("positive" = "accept", "negative" = "reject",   "neutral" = "possible",
     name = "groupy", type = "character")
fnew("positive" = "pass",   "negative" = "fail",     "neutral" = "retest",
     name = "groupz", type = "character")

type     <- c(1, 1, 1, 2, 2, 2, 3, 3, 3)
response <- c("positive", "negative", "neutral",
              "positive", "negative", "neutral",
              "positive", "negative", "neutral")

# Step 1: map type -> format name
respfmt <- fput(type, "typefmt")

# Step 2: apply per-element format
word <- fputc(response, respfmt)

data.frame(type = type, response = response, respfmt = respfmt, word = word)

fclear()

Example 17: Working with Dates and Formats — PUTN

A SAS-style workflow where format names are looked up dynamically per observation:

# Format that maps key codes to date format names
fnew("1" = "date9.", "2" = "mmddyy10.",
     name = "writfmt", type = "numeric")

fnew_date("date9.")
fnew_date("mmddyy10.")

# Input data (R date numbers = days since 1970-01-01)
number <- c(12103, 10899)
key    <- c(1, 2)

# Look up format name per observation
datefmt <- fputn(key, "writfmt")

# Apply per-element date format
date <- fputn(number, datefmt)

data.frame(number = number, key = key, datefmt = datefmt, date = date)

fclear()

Example 18: Import SAS Formats from CNTLOUT CSV

The fimport() function reads a CSV file exported from a SAS format catalogue (PROC FORMAT ... CNTLOUT=):

csv_path <- system.file("extdata", "test_cntlout.csv", package = "ksformat")

imported <- fimport(csv_path)
names(imported)

flist()
fprint()

Use Imported Formats

# Character format (GENDER)
gender_codes <- c("M", "F", NA, "X")
data.frame(
  code = gender_codes,
  label = fputc(gender_codes, "GENDER")
)

# Numeric format (AGEGRP)
ages <- c(5, 17, 18, 45, 65, 100, NA, -1)
data.frame(
  age = ages,
  group = fputn(ages, "AGEGRP")
)

# Numeric format (BMICAT)
bmi_values <- c(15.0, 18.5, 22.3, 25.0, 28.7, 30.0, 35.5)
data.frame(
  bmi = bmi_values,
  category = fputn(bmi_values, "BMICAT")
)

# Invalue (RACEIN)
race_labels <- c("White", "Black", "Asian", "Other")
data.frame(
  label = race_labels,
  code = finputn(race_labels, "RACEIN")
)

Apply to Data Frame

df <- data.frame(
  id = 1:5,
  sex = c("M", "F", "M", NA, "F"),
  age = c(10, 30, 70, NA, 50),
  stringsAsFactors = FALSE
)

gender_fmt <- imported[["GENDER"]]
age_fmt    <- imported[["AGEGRP"]]

fput_df(df, sex = gender_fmt, age = age_fmt, suffix = "_label")

Export Imported Format

cat(fexport(AGEGRP = age_fmt))
cat(fexport(GENDER = gender_fmt))

Selective Import (No Auto-register)

fclear()

manual <- fimport(csv_path, register = FALSE)

# Library should be empty
flist()
fprint()

# Use directly from returned list
fput(c("M", "F"), manual[["GENDER"]])

fclear()

Example 19: Bilingual Format

Expression labels can select between languages at apply-time using an extra argument:

# Single format, language selected via .x1 extra argument
sex_bi <- fnew(
  "M" = "ifelse(.x1 == 'en', 'Male', 'Homme')",
  "F" = "ifelse(.x1 == 'en', 'Female', 'Femme')",
  .missing = "Unknown",
  name = "sex_bi"
)

# .x1 = language code per observation
fput(c("M", "F", "M"), sex_bi, c("en", "fr", "en"))
# -> "Male" "Femme" "Male"

# Alternative: one format per language, selected at apply-time
fnew("M" = "Male",  "F" = "Female",  .missing = "Unknown", name = "sex_en")
fnew("M" = "Homme", "F" = "Femme",   .missing = "Inconnu", name = "sex_fr")

lang <- "fr"
fput(c("M", "F", NA), paste0("sex_", lang))
# -> "Homme" "Femme" "Inconnu"

fclear()

Example 20: Composite Key Lookup with fputk()

fputk() pastes multiple vectors into a composite key before format lookup. This is useful when a format is keyed on the combination of several columns, a common pattern in clinical data (e.g., looking up a visit date by subject + visit number).

# Simulate a Subject Visits (SV) domain
SV <- data.frame(
  USUBJID  = c("SUBJ-001", "SUBJ-001", "SUBJ-001", "SUBJ-002", "SUBJ-002"),
  VISITNUM = c(1, 2, 3, 1, 2),
  SVSTDTC  = c("2025-01-15", "2025-02-20", "2025-03-10",
               "2025-01-18", "2025-02-25"),
  stringsAsFactors = FALSE
)

# Simulate a Questionnaires (QS) domain
QS <- data.frame(
  USUBJID  = c("SUBJ-001", "SUBJ-001", "SUBJ-002", "SUBJ-002", "SUBJ-002"),
  VISITNUM = c(1, 2, 1, 2, 3),
  QSTESTCD = c("SCORE1", "SCORE1", "SCORE1", "SCORE1", "SCORE1"),
  QSSTRESN = c(85, 90, 72, 78, NA),
  stringsAsFactors = FALSE
)

SV
QS

Character lookup (returns character strings)

Register a format keyed on USUBJID|VISITNUM with values being the visit start date (SVSTDTC) as character strings:

# Create composite key -> date string mapping from SV
fnew(
  fmap(paste(SV$USUBJID, SV$VISITNUM, sep = "|"), SV$SVSTDTC),
  .other  = "NOT FOUND",
  name    = "svdtc",
  type    = "character",
  ignore_case = TRUE
)

fprint("svdtc")

Now look up visit dates in the QS domain using fputk():

QS$SVSTDTC <- fputk(QS$USUBJID, QS$VISITNUM, format = "svdtc")
QS
class(QS$SVSTDTC)  # character

fclear()

Native Date lookup (returns Date objects)

Using type = "Date", values are stored as native R Date objects and fput()/fputk() return them directly — no string conversion needed:

# Create composite key -> Date mapping from SV
fnew(
  fmap(
    paste(SV$USUBJID, SV$VISITNUM, sep = "|"),
    as.Date(SV$SVSTDTC, format = "%Y-%m-%d")
  ),
  .other  = NA,
  name    = "svdtn",
  type    = "Date",
  ignore_case = TRUE
)

fprint("svdtn")

QS$SVSTDTC_DT <- fputk(QS$USUBJID, QS$VISITNUM, format = "svdtn")
QS
class(QS$SVSTDTC_DT)  # Date

# Typed NA for unmatched keys (SUBJ-002 Visit 3 not in SV)
is.na(QS$SVSTDTC_DT[5])

# Date arithmetic works directly
QS$SVSTDTC_DT + 7  # add 7 days

fclear()

Example 21: Consistent Data-Driven Formats with fmap()

When building formats from data (e.g., a data frame with 1000+ rows), you need a named vector mapping keys to values. By default, fnew() treats named vectors differently depending on the output type:

• Value types (Date, POSIXct, logical): c(key = value) — natural direction, no reversal.
• Character / numeric: c(Label = "Code") — R convention, names and values are swapped internally.

This inconsistency is confusing for data-driven formats. The fmap() helper solves it: fmap(keys, values) works identically for all types.

Clinical-data example

Suppose we have a demographics dataset and need two lookup formats from the same data — one returning Date objects, one returning character strings:

library(ksformat)

dm <- data.frame(
  USUBJID = c("SUBJ-001", "SUBJ-002", "SUBJ-003"),
  SUBJID  = c("001", "002", "003"),
  RFICDTC = c("2023-03-09T08:45", "2024-08-13T09:53", "2025-06-17T09:03"),
  stringsAsFactors = FALSE
)

# Composite key for both formats
keys <- paste(dm$USUBJID, dm$SUBJID, sep = "|")

Same fmap(keys, values) pattern for both types

Both formats use the identical calling style — fmap(keys, values) where keys are input lookup values and values are output objects:

# Date lookup
fnew(
  fmap(keys, as.Date(dm$RFICDTC, format = "%Y-%m-%d")),
  .other      = NA,
  type        = "Date",
  ignore_case = TRUE,
  name        = "icdtn"
)

# Character lookup — same fmap(keys, values) pattern!
fnew(
  fmap(keys, dm$RFICDTC),
  .other      = "NOT FOUND",
  type        = "character",
  ignore_case = TRUE,
  name        = "icdtc"
)

fprint("icdtn")
fprint("icdtc")

# Both return the expected results
fputk("SUBJ-001", "001", format = "icdtn")
class(fputk("SUBJ-001", "001", format = "icdtn"))

fputk("SUBJ-001", "001", format = "icdtc")
class(fputk("SUBJ-001", "001", format = "icdtc"))

fclear()

No extra parameters needed — fmap() tells fnew() to use the natural direction for all types.

When to use the default (reversal on)

The default auto-reversal preserves the standard R convention where c(Label = "Code") maps Code -> Label. This is natural for hand-written formats:

# These are equivalent — both map "M" -> "Male"
fmt_a <- fnew(c(Male = "M", Female = "F"))
fmt_b <- fnew("M" = "Male", "F" = "Female")

identical(fput(c("M", "F"), fmt_a), fput(c("M", "F"), fmt_b))

fclear()

Summary

| Use case | Style | Reversal | |:--------------------------|:--------------------------------------|:----------| | Data-driven (any type) | fmap(keys, values) | Suppressed | | Hand-written (char/num) | c(Label = "Code") or "Code" = "Label" | Auto (default) | | Value types (Date, etc.)| fmap(keys, values) or setNames(values, keys) | No reversal (default) |

Example 22: Date Lookup via fparse() and fputk()

Examples 20–21 built composite-key formats programmatically with fnew() and fmap(). When the mapping is small and known in advance (e.g., a study-specific visit schedule), you can define the same lookup entirely in text with fparse().

Character date lookup

The simplest approach: store dates as character strings using a regular character format.

fparse(text = '
VALUE svdtc (character, nocase)
  "SUBJ-001|1" = "2025-01-15"
  "SUBJ-001|2" = "2025-02-20"
  "SUBJ-001|3" = "2025-03-10"
  "SUBJ-002|1" = "2025-01-18"
  "SUBJ-002|2" = "2025-02-25"
  .other       = "NOT FOUND"
;
')

fprint("svdtc")

Apply with fputk() to look up visit dates from a questionnaire domain:

QS <- data.frame(
  USUBJID  = c("SUBJ-001", "SUBJ-001", "SUBJ-002", "SUBJ-002", "SUBJ-002"),
  VISITNUM = c(1, 2, 1, 2, 3),
  QSSTRESN = c(85, 90, 72, 78, NA),
  stringsAsFactors = FALSE
)

QS$SVSTDTC <- fputk(QS$USUBJID, QS$VISITNUM, format = "svdtc")
QS

fclear()

Native Date lookup

Use the Date value type with format: to store dates as native R Date objects. The format: parameter tells fparse() how to parse the date strings in the text block:

fparse(text = '
VALUE svdtn (Date, format: %Y-%m-%d, nocase)
  "SUBJ-001|1" = "2025-01-15"
  "SUBJ-001|2" = "2025-02-20"
  "SUBJ-001|3" = "2025-03-10"
  "SUBJ-002|1" = "2025-01-18"
  "SUBJ-002|2" = "2025-02-25"
;
')

fprint("svdtn")

Now fputk() returns real Date objects — arithmetic and comparison work directly:

QS$SVSTDTC_DT <- fputk(QS$USUBJID, QS$VISITNUM, format = "svdtn")
QS

class(QS$SVSTDTC_DT)         # Date
is.na(QS$SVSTDTC_DT[5])      # TRUE — no match for SUBJ-002 Visit 3

# Date arithmetic works directly
QS$SVSTDTC_DT + 7

Round-trip: export and re-import

Formats created with fparse() can be exported back to text with fexport() and re-parsed — useful for version-controlled format definitions:

fmt_obj <- format_get("svdtn")
txt <- fexport(svdtn = fmt_obj)
cat(txt)

# Re-parse the exported text
fclear()
fparse(text = txt)

# Verify it still works
fputk("SUBJ-001", 2, format = "svdtn")

fclear()

Example 23: Inspecting Range Rules with franges()

franges() extracts all range-based mappings from a format and returns them as a tidy data.frame — useful for auditing, documentation, or downstream processing.

fparse(text = '
VALUE age (numeric)
  [0, 18)    = "Child"
  [18, 65)   = "Adult"
  [65, HIGH] = "Senior"
  .missing   = "Unknown"
;
')

franges("age")

You can use the result like any data frame — filter, display, or feed into further calculations:

df <- franges("age")

# Which ranges have a finite upper bound?
df[is.finite(df$high), ]

franges() silently excludes discrete entries (.missing, .other, plain string keys) — only range rows appear. It returns an empty data.frame with the same columns when the format contains no ranges.

fnew("M" = "Male", "F" = "Female", .missing = "Unknown", name = "sex")
franges("sex")   # 0 rows

fclear()

Example 24: Reverse Range Lookup with fmap_to_ranges()

When a range format stores numeric codes as its labels (e.g. visit windows coded as weeks), fmap_to_ranges() turns a vector of those codes back into the original [low, high] bounds — one row per input value.

fparse(text = '
VALUE visit_ther (numeric)
  [LOW,  1] =  0
  [ 8, 22] =  2
  [22, 36] =  4
  [37, 50] =  6
  [51, 63] =  8
  [64, 78] = 10
  [79, 91] = 12
;
')

coded_weeks <- c(0, 2, 4, 6, 8, 10, 12)
fmap_to_ranges(coded_weeks, "visit_ther")

Unmatched values produce NA rows, making it safe to pass arbitrary vectors:

fmap_to_ranges(c(2, 99, 4), "visit_ther")

fclear()

Example 25: Date Range Bucketing

date_range and datetime_range formats bucket Date or POSIXct input into character labels using ISO date/datetime interval bounds. They reuse the same range-table engine as numeric ranges, so the findInterval() fast path is active for sorted, disjoint buckets.

Fiscal-year bucketing

fnew(
  "2023-01-01,2024-01-01,TRUE,FALSE" = "FY23",
  "2024-01-01,2025-01-01,TRUE,FALSE" = "FY24",
  "2025-01-01,2026-01-01,TRUE,FALSE" = "FY25",
  type = "date_range",
  name = "fiscal_year"
)

dates <- as.Date(c("2023-06-15", "2024-03-01", "2024-12-31",
                   "2025-07-04", "2022-01-01", NA))

data.frame(
  date  = dates,
  fy    = fput(dates, "fiscal_year")
)

Define from text with fparse()

fparse(text = '
VALUE quarter (date_range)
  [2024-01-01, 2024-04-01) = "Q1-2024"
  [2024-04-01, 2024-07-01) = "Q2-2024"
  [2024-07-01, 2024-10-01) = "Q3-2024"
  [2024-10-01, 2025-01-01) = "Q4-2024"
  .other                   = "Outside 2024"
;
')

sample_dates <- as.Date(c("2024-02-14", "2024-05-20", "2024-08-08",
                          "2024-11-30", "2025-03-01"))

data.frame(
  date    = sample_dates,
  quarter = fput(sample_dates, "quarter")
)

LOW / HIGH open-ended bounds

LOW and HIGH represent $-\infty$ and $+\infty$ — any date before or after a cutpoint falls in the open arm.

fparse(text = '
VALUE era (date_range)
  [LOW,        2000-01-01) = "Pre-2000"
  [2000-01-01, 2010-01-01) = "2000s"
  [2010-01-01, 2020-01-01) = "2010s"
  [2020-01-01, HIGH]       = "2020+"
;
')

event_dates <- as.Date(c("1985-07-04", "2005-12-25",
                         "2015-06-01", "2023-11-11"))

data.frame(
  date = event_dates,
  era  = fput(event_dates, "era")
)

Export and roundtrip

Formats export with ISO date bounds and re-parse without loss:

q_obj <- format_get("quarter")
cat(fexport(quarter = q_obj))

# Re-parse the exported text
txt <- fexport(quarter = q_obj)
fclear()
fparse(text = txt)

fput(as.Date(c("2024-02-14", "2024-08-08")), "quarter")

Overlapping buckets with multilabel and fput_all()

fparse(text = '
VALUE study_window (date_range, multilabel)
  [2024-01-01, 2024-07-01) = "First Half"
  [2024-04-01, 2024-10-01) = "Mid-Year"
  [2024-07-01, 2025-01-01) = "Second Half"
;
')

checkup_dates <- as.Date(c("2024-02-15", "2024-05-20", "2024-09-01"))
all_windows   <- fput_all(checkup_dates, "study_window")

for (i in seq_along(checkup_dates)) {
  cat(format(checkup_dates[i]), "->",
      paste(all_windows[[i]], collapse = " | "), "\n")
}

Auto-detection of type

When no explicit type is given, fparse() infers date_range from ISO date bounds and datetime_range when bounds include a time component:

fparse(text = '
VALUE auto_fy
  [2024-01-01, 2025-01-01) = "2024"
;

VALUE auto_shift
  [2024-01-15 08:00, 2024-01-15 16:00) = "Day shift"
;
')

cat("auto_fy type   :", format_get("auto_fy")$type, "\n")
cat("auto_shift type:", format_get("auto_shift")$type, "\n")

Datetime range bucketing

datetime_range works identically to date_range but matches against POSIXct values. Bounds are expressed as YYYY-MM-DD HH:MM[:SS] strings.

fparse(text = '
VALUE shift (datetime_range)
  [2024-01-15 00:00, 2024-01-15 08:00) = "Night"
  [2024-01-15 08:00, 2024-01-15 16:00) = "Day"
  [2024-01-15 16:00, 2024-01-16 00:00) = "Evening"
;
')

timestamps <- as.POSIXct(
  c("2024-01-15 03:22:00", "2024-01-15 11:45:00",
    "2024-01-15 19:00:00"),
  tz = "UTC"
)

data.frame(
  ts    = format(timestamps, tz = "UTC"),
  shift = fput(timestamps, "shift")
)

fclear()

Example 26: Stratified Range Lookup with fputk()

The stratified_range type combines a discrete stratum (such as a study arm, subject id, or any composite key) with a numeric / Date / POSIXct range. Each stratum has its own bucket boundaries, and fputk() dispatches to the right bucket for each row.

Programmatic construction with fmap_strata()

visits <- fmap_strata(
  stratum = c("ARM_A", "ARM_A", "ARM_A", "ARM_B", "ARM_B"),
  low     = c(0,        7,       28,      0,       14),
  high    = c(7,        28,      Inf,     14,      Inf),
  label   = c("Baseline", "Wk1-3", "Wk4+", "Baseline", "Wk2+"),
  inc_high = c(FALSE, FALSE, TRUE, FALSE, TRUE)
)
fnew(visits, type = "stratified_range",
     ".other|ARM_A" = "A_outside",
     .other = "outside_window",
     name = "vw")

df <- data.frame(
  arm = c("ARM_A", "ARM_A", "ARM_B", "ARM_B", "ARM_C"),
  day = c(3,        35,      5,       40,      10)
)
df$visit <- fputk(df$arm, df$day, format = "vw")
df

Text definition with fparse()

fparse(text = '
VALUE vw_text (stratified_range, range_subtype: numeric)
  "ARM_A"|[0, 7)    = "Baseline"
  "ARM_A"|[7, 28)   = "Wk1-3"
  "ARM_A"|[28, HIGH]= "Wk4+"
  "ARM_B"|[0, 14)   = "Baseline"
  "ARM_B"|[14, HIGH]= "Wk2+"
  ".other|ARM_A"    = "A_outside"
  .other            = "outside_window"
  ;
')
fputk(df$arm, df$day, format = "vw_text")

Date subtype: per-subject windows

windows <- fmap_strata(
  stratum = c("S001", "S001", "S002", "S002"),
  low     = as.Date(c("2024-01-01", "2024-01-15",
                       "2024-02-01", "2024-02-20")),
  high    = as.Date(c("2024-01-15", "2024-02-01",
                       "2024-02-20", "2024-03-10")),
  label   = c("Screen", "Treat", "Screen", "Treat")
)
fnew(windows, type = "stratified_range", range_subtype = "date",
     .other = "off-window", name = "win")

subj   <- c("S001", "S001", "S002", "S002", "S003")
visits <- as.Date(c("2024-01-05", "2024-01-20",
                     "2024-02-10", "2024-03-01", "2024-01-01"))
data.frame(
  subj  = subj,
  date  = visits,
  phase = fputk(subj, visits, format = "win")
)

Roundtrip via fexport() / fparse()

txt <- fexport(format_get("vw"))
cat(txt, "\n")
fclear()
fparse(text = txt)
fputk(df$arm, df$day, format = "vw")

fclear()

Example 27: Plain Range Lookup with fmap_ranges()

For non-stratified numeric / Date ranges, fmap_ranges() saves you from hand-crafting canonical \code{"low,high,inc_low,inc_high"} keys.

age_groups <- fmap_ranges(
  low   = c(0, 18, 65),
  high  = c(18, 65, Inf),
  label = c("Child", "Adult", "Senior"),
  inc_high = c(FALSE, FALSE, TRUE)
)
fnew(age_groups, type = "numeric", name = "ag")
fput(c(5, 25, 90), "ag")
fclear()

Example 28: Composite Key Lookup with NA Components (na_as_string)

When building a format from data using fmap(paste(..., sep = "|"), values), base R's paste() converts any NA component to the literal string "NA". The resulting composite key is therefore "CAT|TEST|NA", not a missing value.

By default, fputk() restores NA_character_ wherever any component is NA before the lookup — so the key "CAT|TEST|NA" is never reached and the row falls through to .other / .missing.

Setting na_as_string = TRUE keeps paste()'s literal "NA", making the round-trip consistent.

Clinical example — LB parameter derivation

A common ADaM task: derive PARAMCD from a combination of LBCAT, LBSPEC, LBTESTCD, and LBSTRESU, where some rows have LBSTRESU = NA (dimensionless tests such as INR).

# Source lab mapping (as received from a specification)
lb_map <- data.frame(
  LBCAT    = c("BLOOD CHEMISTRY", "COAGULOGRAM", "COAGULATION PANEL", "COAGULOGRAM"),
  LBSPEC   = c("BLOOD",           "BLOOD",        "BLOOD",             "BLOOD"),
  LBTESTCD = c("ALB",             "FIBRINO",      "INR",               "INR"),
  LBSTRESU = c("g/L",             "g/L",           NA,                  NA),
  PARAMCD  = c("ALB",             "FIBRINO",       "INR",               "INR"),
  stringsAsFactors = FALSE
)
lb_map

Build the format with fmap(paste(...), PARAMCD).
paste() converts NA in LBSTRESU to "NA", so the stored keys for INR rows are "COAGULATION PANEL|BLOOD|INR|NA" and "COAGULOGRAM|BLOOD|INR|NA".

with(lb_map,
  fmap(paste(LBCAT, LBSPEC, LBTESTCD, LBSTRESU, sep = "|"), PARAMCD)
) |>
  fnew(ignore_case = TRUE, .other = NA,
       type = "character", name = "lb_param")

fprint("lb_param")

Now apply the format.
With the default na_as_string = FALSE, the INR rows get NA (no match):

lb_map$PARAMCD_default <- with(lb_map,
  fputk(LBCAT, LBSPEC, LBTESTCD, LBSTRESU, format = "lb_param")
)
lb_map[, c("LBTESTCD", "LBSTRESU", "PARAMCD", "PARAMCD_default")]

With na_as_string = TRUE, paste() also converts the lookup-side NA to "NA", so the keys match:

lb_map$PARAMCD_back <- with(lb_map,
  fputk(LBCAT, LBSPEC, LBTESTCD, LBSTRESU,
        format = "lb_param", na_as_string = TRUE)
)
lb_map[, c("LBTESTCD", "LBSTRESU", "PARAMCD", "PARAMCD_back")]

fclear()

The rule of thumb

Use na_as_string = TRUE whenever the format was built with
fmap(paste(...), values) and any key column can contain NA.

If the format keys were set by hand (fnew("CAT|TEST|g/L" = "ALB", ...)), NA components should still go through .missing — keep the default na_as_string = FALSE.

Example 29: Composite Label Invalue Lookup with finputk()

finputk() is the invalue-side mirror of fputk(): it pastes multiple columns into a composite label and reverse-looks it up in a ks_invalue format. The same na_as_string argument applies.

Basic usage

# Build an INVALUE from two-column composite labels
finput(
  fmap(paste(c("BLOOD CHEMISTRY", "COAGULOGRAM", "COAGULATION PANEL"),
             c("ALB",             "FIBRINO",      "INR"),
             sep = "|"),
       c(1L, 2L, 3L)),
  target_type = "integer",
  name = "lb_code_inv"
)

# Reverse lookup: two separate columns → integer code
cat_vec  <- c("BLOOD CHEMISTRY", "COAGULOGRAM", "COAGULATION PANEL", "OTHER")
test_vec <- c("ALB",              "FIBRINO",     "INR",               "X")

finputk(cat_vec, test_vec, invalue_name = "lb_code_inv")
# BLOOD CHEMISTRY|ALB → 1, COAGULOGRAM|FIBRINO → 2,
# COAGULATION PANEL|INR → 3, OTHER|X → NA (no match → missing_value)
fclear()

With NA components (na_as_string = TRUE)

When the INVALUE was built from data containing NA columns, use na_as_string = TRUE on both the build side (fmap(paste(...), ...)) and the lookup side (finputk(..., na_as_string = TRUE)).

# INVALUE where LBSTRESU can be NA (like INR)
finput(
  fmap(
    paste(lb_map$LBCAT, lb_map$LBTESTCD, lb_map$LBSTRESU, sep = "|"),
    seq_len(nrow(lb_map))
  ),
  target_type = "integer",
  name = "lb_row_inv"
)

# Reconstruct lb_map row indices — works even when LBSTRESU is NA
finputk(lb_map$LBCAT, lb_map$LBTESTCD, lb_map$LBSTRESU,
        invalue_name = "lb_row_inv", na_as_string = TRUE)

fclear()

The output type is always determined by the stored invalue's target_type (here integer). For character output create the invalue with target_type = "character" and finputk() returns a character vector.

ksformat documentation built on May 21, 2026, 9:07 a.m.