Finds the maximal stretches of consecutive NA in each column,
with where each begins and how long it is.
Row order carries meaning in a capsule far more often than people allow for – a time series, an ordered export, a paginated download – and a long unbroken run of missingness means something different from the same count scattered about. A run says an instrument was down, a page failed to fetch, or a period was never collected; scattered gaps say individual records failed. The rate cannot distinguish them.
Value
A data frame of class bricklayer_runs with column
(character), start and end (integer row numbers, inclusive) and
length (integer), longest first and then by column name. Zero rows
when there are no qualifying runs.
Details
Each column's missingness indicator is.na(x) is run-length
encoded with rle(); every run of TRUE with length at least
minrun is reported. Rows are taken in their current order, so sort
the data by time (or by download order) first. A list column counts an
element as missing only when is.na() says so. Errors: data not a
data frame or without columns, or with duplicated or empty names;
minrun not a positive whole number.
See also
misspat() for which
columns are missing together,
misssmry() for the rates.
Examples
# One long outage and two isolated gaps, with the same total count.
df <- data.frame(
outage = c(1, 2, NA, NA, NA, NA, 7, 8),
scattered = c(1, NA, 3, 4, NA, 6, NA, NA)
)
sum(is.na(df$outage)) == sum(is.na(df$scattered))
#> [1] TRUE
missruns(df)
#> ╭── Runs of consecutive missing values ───────────────────────────────────╮
#> │ Runs 2 │
#> │ Columns with runs 2 │
#> │ Longest run 4 │
#> │ │
#> │ column start end length │
#> │ --------- ----- --- ------ │
#> │ outage 3 6 4 │
#> │ scattered 7 8 2 │
#> │ │
#> │ Long runs of missing rows usually mean a source stopped reporting for │
#> │ a while, not values missing at random. │
#> ╰─────────────────────────────────────────────────────────────────────────╯
# Isolated gaps too, by lowering the threshold.
missruns(df, minrun = 1)
#> ╭── Runs of consecutive missing values ───────────────────────────────────╮
#> │ Runs 4 │
#> │ Columns with runs 2 │
#> │ Longest run 4 │
#> │ │
#> │ column start end length │
#> │ --------- ----- --- ------ │
#> │ outage 3 6 4 │
#> │ scattered 7 8 2 │
#> │ scattered 2 2 1 │
#> │ scattered 5 5 1 │
#> │ │
#> │ Long runs of missing rows usually mean a source stopped reporting for │
#> │ a while, not values missing at random. │
#> ╰─────────────────────────────────────────────────────────────────────────╯
# A complete column has no runs.
missruns(data.frame(x = 1:5))
#> ╭── Runs of consecutive missing values ──╮
#> │ Runs 0 │
#> │ Columns with runs 0 │
#> │ │
#> │ (none) │
#> ╰────────────────────────────────────────╯
try(missruns(df, minrun = 0))
#> Error : `minrun` must be a single positive integer