Collapses a parsed address into a single string that two spellings of the same address share, which is what joining or deduplicating two address lists needs. Components are folded to an accent- and case-insensitive form and punctuation is dropped, so `St. John's` and `SAINT JOHNS` key alike, and the fields are laid out broad to narrow – province, municipality, street, civic number – so sorting the keys clusters a street together.
Normalizing is what does the real work here; the key only makes the result joinable. Pass `con` (or normalize with it first) whenever the input is messy: the gazetteer is what turns a misspelled street into NAR's own spelling, and two lists cannot key alike on a name only one of them got right.
Arguments
- x
Either a data frame of parsed components, as returned by [normalize_address()], or a character vector of address strings to normalize first.
- unit
Include the unit number in the key? The default `FALSE` keys a *building*, so every suite in a tower collapses to one key. Set it to `TRUE` to key a tenant instead – at the cost that the unit is the least reliably parsed component, so rows that agree on everything else will split whenever one of them wrote its suite somewhere the parser did not find it.
- sep
The separator between fields. It only has to be a character the components cannot contain; the default is fine unless a downstream tool treats `|` specially.
- known, con
Passed to [normalize_address()], and only allowed when `x` is a character vector – a data frame has already been parsed.
Value
A character vector, one element per row of `x`, and `NA` for any row with no street name. Those are rows nothing could be keyed from, and they are `NA` rather than an empty key so they cannot all collapse onto each other. Note that `dplyr`'s joins match `NA` to `NA` by default, so filter them out or pass `na_matches = "never"`.
See also
[format_address()] for the human-readable form, [normalize_address()] for the components themselves.
Examples
# Five spellings of one building, one key.
address_key(c("1055 W Georgia St, Vancouver, BC",
"1055 West Georgia Street, Vancouver, British Columbia",
"Suite 1500 - 1055 W Georgia St, Vancouver BC",
"#1500-1055 west georgia st., vancouver, b.c.",
"1055 WEST GEORGIA ST, VANCOUVER, BC V6E 3P3"))
#> [1] "BC|VANCOUVER|GEORGIA|ST|W|1055|" "BC|VANCOUVER|GEORGIA|ST|W|1055|"
#> [3] "BC|VANCOUVER|GEORGIA|ST|W|1055|" "BC|VANCOUVER|GEORGIA|ST|W|1055|"
#> [5] "BC|VANCOUVER|GEORGIA|ST|W|1055|"
# Keying the tenant instead of the building keeps the suites apart.
address_key(c("Suite 1500 - 1055 W Georgia St, Vancouver BC",
"Suite 800 - 1055 W Georgia St, Vancouver BC"), unit = TRUE)
#> [1] "BC|VANCOUVER|GEORGIA|ST|W|1055||1500"
#> [2] "BC|VANCOUVER|GEORGIA|ST|W|1055||800"