ateucher · ateucher · Jul 12, 2019 · May 3, 2019 · May 3, 2019 · Jul 11, 2019
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -1,7 +1,7 @@
 Package: lutz
 Type: Package
 Title: Look Up Time Zones of Point Coordinates
-Version: 0.2.0
+Version: 0.2.0.999
 Authors@R: person("Andy", "Teucher", email = "andy.teucher@gmail.com", role = c("aut", "cre"))
 Description: Input latitude and longitude values or an 'sf/sfc' POINT 
     object and get back the timezone in which they exist. Two methods are implemented. 
@@ -28,5 +28,5 @@ Suggests:
     covr
 Encoding: UTF-8
 LazyData: true
-RoxygenNote: 6.0.1
+RoxygenNote: 6.1.1
 Roxygen: list(markdown = TRUE)
diff --git a/NEWS.md b/NEWS.md
@@ -1,3 +1,8 @@
+# lutz 0.2.0.999
+
+* Upgraded timezone map to 2019b
+* Dealt with areas with overlapping timezones (#2)
+
 # lutz 0.2.0
 
 * Added `method = "accurate"` to do a slower, but more accurate lookup.

diff --git a/R/sysdata.rda b/R/sysdata.rda
diff --git a/R/tz_lookup.R b/R/tz_lookup.R
@@ -6,6 +6,13 @@
 #' incorrect timezone. If accuracy is more important than speed, use
 #' `method = "accurate"`.
 #'
+#' Note that there are some regions in the world where a single point can land in
+#' two different overlapping timezones. The `"accurate"` method includes these,
+#' and when they are encountered they are concatenated in a single string,
+#' separated by a semicolon.
+#' The data used in the `"fast"` method does not include overlapping timezones
+#' at this time.
+#'
 #' @param x either an `sfc` or `sf` points or `SpatialPoints(DataFrame)` object
 #' @param crs the coordinate reference system: integer with the EPSG code, or character with proj4string.
 #' If not specified (i.e., `NULL`) and `x` has no existing `crs`, EPSG: 4326 is assumed (lat/long).
@@ -36,7 +43,7 @@ tz_lookup <- function(x, crs = NULL, method = "fast", warn = TRUE) {
   switch(method,
          fast = tz_lookup_fast(x, crs, warn),
          accurate = tz_lookup_accurate(x, crs),
-         stop("method mst be one of 'fast' or 'accurate'", call. = FALSE))
+         stop("method must be one of 'fast' or 'accurate'", call. = FALSE))
 
 }
 
@@ -83,9 +90,9 @@ tz_lookup_coords <- function(lat, lon, method = "fast", warn = TRUE) {
   check_coords(lat, lon)
 
   switch(method,
-    fast = tz_lookup_coords_fast(lat, lon, warn),
-    accurate = tz_lookup_coords_accurate(lat, lon),
-    stop("method mst be one of 'fast' or 'accurate'", call. = FALSE)
+         fast = tz_lookup_coords_fast(lat, lon, warn),
+         accurate = tz_lookup_coords_accurate(lat, lon),
+         stop("method must be one of 'fast' or 'accurate'", call. = FALSE)
   )
 }
 
@@ -107,8 +114,24 @@ tz_lookup_accurate <- function(x, crs = NULL) {
 
 tz_lookup_accurate.sf <- function(x, crs = NULL) {
   x <- fix_sf(x, crs)
-  x <- suppressMessages(sf::st_join(x, tz_sf))
-  ret <- x$tzid
+  # Add a unique id so we can deal with any duplicates resulting
+  # from overlapping timezones
+  x$lutzid <- seq_len(nrow(x))
+  x_tz <- suppressMessages(sf::st_set_geometry(sf::st_join(x, tz_sf), NULL))
+
+  # group x by lutzid and concatenate multiple timezones with ;
+  if (nrow(x_tz) > nrow(x)) {
+    warning("Some points are in areas with more than one timezone defined.",
+            "These are often disputed areas and should be treated with care.")
+
+    ret <- stats::aggregate(x_tz, list(x_tz$lutzid), function(x) {
+      if (length(x) == 1) return(x)
+      x <- paste(x, collapse = "; ")
+    }, drop = FALSE)[["tzid"]]
+
+  } else {
+    ret <- x_tz$tzid
+  }
 
   # If any are NA, try to fill in with V8-based tzlookup
   nas <- which(is.na(ret))
@@ -120,7 +143,7 @@ tz_lookup_accurate.sf <- function(x, crs = NULL) {
 }
 
 tz_lookup_accurate.sfc <- function(x, crs = NULL) {
-  x_sf <- sf::st_sf(id = seq_len(length(x)), geom = x)
+  x_sf <- sf::st_sf(geom = x)
   tz_lookup_accurate(x_sf, crs = crs)
 }
 

diff --git a/README.Rmd b/README.Rmd
@@ -12,7 +12,13 @@ knitr::opts_chunk$set(
 )
 ```
 
-[![Travis-CI Build Status](https://travis-ci.org/ateucher/lutz.svg?branch=master)](https://travis-ci.org/ateucher/lutz) [![AppVeyor Build Status](https://ci.appveyor.com/api/projects/status/github/ateucher/lutz?branch=master&svg=true)](https://ci.appveyor.com/project/ateucher/lutz) [![Coverage Status](https://img.shields.io/codecov/c/github/ateucher/lutz/master.svg)](https://codecov.io/github/ateucher/lutz?branch=master)
+<!-- badges: start -->
+[![Travis-CI Build Status](https://travis-ci.org/ateucher/lutz.svg?branch=master)](https://travis-ci.org/ateucher/lutz)
+[![AppVeyor Build Status](https://ci.appveyor.com/api/projects/status/github/ateucher/lutz?branch=master&svg=true)](https://ci.appveyor.com/project/ateucher/lutz)
+[![Coverage Status](https://img.shields.io/codecov/c/github/ateucher/lutz/master.svg)](https://codecov.io/github/ateucher/lutz?branch=master)
+[![CRAN status](https://www.r-pkg.org/badges/version/lutz)](https://cran.r-project.org/package=lutz)
+[![CRAN downloads](https://cranlogs.r-pkg.org/badges/lutz)](https://cran.r-project.org/package=lutz)
+<!-- badges: end -->
 
 # lutz (look up timezones)
 
@@ -91,6 +97,10 @@ ggplot(cbind(as.data.frame(coordinates(state_centers_sp)), tz = state_centers_sp
   theme_minimal()
 ```
 
+Note that there are some regions in the world where a single point can land in 
+two different overlapping timezones. The `"accurate"` method [includes these](https://github.com/evansiroky/timezone-boundary-builder/releases/tag/2018g), 
+however the method used in the `"fast"` does not include overlapping timezones
+([at least for now](https://github.com/darkskyapp/tz-lookup/issues/34)).
 
 We can compare the accuracy of both methods to the high-resolution timezone map
 provided by https://github.com/evansiroky/timezone-boundary-builder. This is the
@@ -99,7 +109,7 @@ simplified by about 80% to be small enough to fit in the package.
 
 ```{r eval=FALSE}
 ## Get the full timezone geojson from https://github.com/evansiroky/timezone-boundary-builder
-download.file("https://github.com/evansiroky/timezone-boundary-builder/releases/download/2018d/timezones-with-oceans.geojson.zip",
+download.file("https://github.com/evansiroky/timezone-boundary-builder/releases/download/2019a/timezones-with-oceans.geojson.zip",
                 destfile = "tz.zip")
 unzip("tz.zip", exdir = "data-raw/dist/")
 ```
@@ -108,19 +118,25 @@ unzip("tz.zip", exdir = "data-raw/dist/")
 ```{r eval=file.exists("data-raw/dist/combined-with-oceans.json")}
 library(lutz)
 library(sf)
-library(rmapshaper)
 library(purrr)
+library(dplyr)
 
 tz_full <- read_sf("data-raw/dist/combined-with-oceans.json")
 # Create a data frame of 500000 lat/long pairs:
 set.seed(1)
 n <- 500000
-ll <- data.frame(lat = runif(n, -90, 90), lon = runif(n, -180, 180))
+ll <- data.frame(id = seq(n), lat = runif(n, -90, 90), lon = runif(n, -180, 180))
 ll_sf <- st_as_sf(ll, coords = c("lon", "lat"), crs = 4326)
 
 # Overlay those points with the full high-resolution timezone map:
 ref_ll_tz <- sf::st_join(ll_sf, tz_full)
 
+# Combine those that had overlapping timezones
+ref_ll_tz <- ref_ll_tz %>% 
+  st_set_geometry(NULL) %>% 
+  group_by(id) %>% 
+  summarize(tzid = paste(tzid, collapse = "; "))
+
 # run tz_lookup with both `"fast"` and `"accurate"` methods and compare with 
 # the timezones looked up with the high-resolution map:
 tests <- map_df(c("fast", "accurate"), ~ {

diff --git a/README.md b/README.md
@@ -1,12 +1,19 @@
 
 <!-- README.md is generated from README.Rmd. Please edit that file -->
 
+<!-- badges: start -->
+
 [![Travis-CI Build
 Status](https://travis-ci.org/ateucher/lutz.svg?branch=master)](https://travis-ci.org/ateucher/lutz)
 [![AppVeyor Build
 Status](https://ci.appveyor.com/api/projects/status/github/ateucher/lutz?branch=master&svg=true)](https://ci.appveyor.com/project/ateucher/lutz)
 [![Coverage
 Status](https://img.shields.io/codecov/c/github/ateucher/lutz/master.svg)](https://codecov.io/github/ateucher/lutz?branch=master)
+[![CRAN
+status](https://www.r-pkg.org/badges/version/lutz)](https://cran.r-project.org/package=lutz)
+[![CRAN
+downloads](https://cranlogs.r-pkg.org/badges/lutz)](https://cran.r-project.org/package=lutz)
+<!-- badges: end -->
 
 # lutz (look up timezones)
 
@@ -99,36 +106,58 @@ ggplot(cbind(as.data.frame(coordinates(state_centers_sp)), tz = state_centers_sp
 
 ![](tools/readme/unnamed-chunk-5-1.png)<!-- -->
 
+Note that there are some regions in the world where a single point can
+land in two different overlapping timezones. The `"accurate"` method
+[includes
+these](https://github.com/evansiroky/timezone-boundary-builder/releases/tag/2018g),
+however the method used in the `"fast"` does not include overlapping
+timezones ([at least for
+now](https://github.com/darkskyapp/tz-lookup/issues/34)).
+
 We can compare the accuracy of both methods to the high-resolution
 timezone map provided by
 <https://github.com/evansiroky/timezone-boundary-builder>. This is the
 map that is used by `lutz` for the `"accurate"` method, but in `lutz` it
-is simplified by about 80% to be small enough to fit in the
-package.
+is simplified by about 80% to be small enough to fit in the package.
 
 ``` r
 ## Get the full timezone geojson from https://github.com/evansiroky/timezone-boundary-builder
-download.file("https://github.com/evansiroky/timezone-boundary-builder/releases/download/2018d/timezones-with-oceans.geojson.zip",
+download.file("https://github.com/evansiroky/timezone-boundary-builder/releases/download/2019a/timezones-with-oceans.geojson.zip",
                 destfile = "tz.zip")
 unzip("tz.zip", exdir = "data-raw/dist/")
 ```
 
 ``` r
 library(lutz)
 library(sf)
-library(rmapshaper)
 library(purrr)
+library(dplyr)
+#> 
+#> Attaching package: 'dplyr'
+#> The following objects are masked from 'package:stats':
+#> 
+#>     filter, lag
+#> The following objects are masked from 'package:base':
+#> 
+#>     intersect, setdiff, setequal, union
 
 tz_full <- read_sf("data-raw/dist/combined-with-oceans.json")
 # Create a data frame of 500000 lat/long pairs:
 set.seed(1)
 n <- 500000
-ll <- data.frame(lat = runif(n, -90, 90), lon = runif(n, -180, 180))
+ll <- data.frame(id = seq(n), lat = runif(n, -90, 90), lon = runif(n, -180, 180))
 ll_sf <- st_as_sf(ll, coords = c("lon", "lat"), crs = 4326)
 
 # Overlay those points with the full high-resolution timezone map:
 ref_ll_tz <- sf::st_join(ll_sf, tz_full)
 #> although coordinates are longitude/latitude, st_intersects assumes that they are planar
+#> although coordinates are longitude/latitude, st_intersects assumes that they are planar
+
+# Combine those that had overlapping timezones
+ref_ll_tz <- ref_ll_tz %>% 
+  st_set_geometry(NULL) %>% 
+  group_by(id) %>% 
+  summarize(tzid = paste(tzid, collapse = "; "))
 
 # run tz_lookup with both `"fast"` and `"accurate"` methods and compare with 
 # the timezones looked up with the high-resolution map:
@@ -147,6 +176,9 @@ tests <- map_df(c("fast", "accurate"), ~ {
     fun_nas = sum(is.na(test_ll_tz))
     )
 })
+#> Warning in tz_lookup_accurate.sf(x, crs): Some points are in areas with
+#> more than one timezone defined.These are often disputed areas and should be
+#> treated with care.
 ```
 
 ``` r
@@ -155,5 +187,5 @@ knitr::kable(tests)
 
 | method   |   time | matches | mismatches | accuracy | ref\_nas | fun\_nas |
 | :------- | -----: | ------: | ---------: | -------: | -------: | -------: |
-| fast     |  2.721 |  384735 |     115265 | 0.769470 |        0 |        0 |
-| accurate | 47.587 |  499956 |         44 | 0.999912 |        0 |        0 |
+| fast     |  2.453 |  371953 |     128047 | 0.743906 |        0 |        0 |
+| accurate | 26.328 |  499949 |         51 | 0.999898 |        0 |        0 |
diff --git a/data-raw/tz_sf.R b/data-raw/tz_sf.R
@@ -1,11 +1,13 @@
 library(sf)
+library(lwgeom)
 library(rmapshaper)
 # The data is from https://github.com/evansiroky/timezone-boundary-builder
 # And released under the Open Data Commons Open Database License (ODbL)
-download.file("https://github.com/evansiroky/timezone-boundary-builder/releases/download/2018d/timezones-with-oceans.geojson.zip",
+download.file("https://github.com/evansiroky/timezone-boundary-builder/releases/download/2019b/timezones-with-oceans.geojson.zip",
               destfile = "data-raw/tz.zip")
 unzip("data-raw/tz.zip", exdir = "data-raw", overwrite = TRUE)
 tz_full <- read_sf(list.files("data-raw", recursive = TRUE, pattern = "oceans\\.json$", full.names = TRUE))
-tz_sf <- ms_simplify(tz_full, keep = 0.17, keep_shapes = TRUE, explode = TRUE)
+tz_sf <- ms_simplify(tz_full, keep = 0.135, keep_shapes = TRUE, explode = TRUE) %>%
+  st_make_valid()
 
-devtools::use_data(tz_sf, compress = "xz", internal = TRUE, overwrite = TRUE)
+usethis::use_data(tz_sf, compress = "xz", internal = TRUE, overwrite = TRUE)
diff --git a/inst/tz-lookup/README.md b/inst/tz-lookup/README.md
@@ -2,9 +2,9 @@
 
 tz-lookup is licensed CC0.
 
-Find the latest release (v6.1.8) at https://github.com/darkskyapp/tz-lookup/releases
+Find the latest release (v6.1.21) at https://github.com/darkskyapp/tz-lookup/releases
 
 ```
 cd inst/tz-lookup
-wget -q https://raw.githubusercontent.com/darkskyapp/tz-lookup/v6.1.8/tz.js -O tz.js
+wget -q https://raw.githubusercontent.com/darkskyapp/tz-lookup/v6.1.21/tz.js -O tz.js
 ```
diff --git a/inst/tz-lookup/tz.js b/inst/tz-lookup/tz.js
diff --git a/man/tz_lookup.Rd b/man/tz_lookup.Rd
diff --git a/tests/testthat/test-tz_lookup.R b/tests/testthat/test-tz_lookup.R
@@ -16,9 +16,9 @@ test_that("warn works with tz_lookup", {
 
 test_that("errors when method is not one of fast, accurate", {
   expect_error(tz_lookup_coords(70,30, method = "great"),
-               "method mst be one of 'fast' or 'accurate'")
+               "method must be one of 'fast' or 'accurate'")
   expect_error(tz_lookup(sf::st_sfc(sf::st_point(c(1,1))), method = "great"),
-               "method mst be one of 'fast' or 'accurate'")
+               "method must be one of 'fast' or 'accurate'")
 })
 
 test_that("tz_lookup_coords works", {

diff --git a/tests/testthat/test-tz_lookup2.R b/tests/testthat/test-tz_lookup2.R
@@ -51,3 +51,18 @@ test_that("tz_lookup.SpatialPoints works", {
                "Etc/GMT+9")
   expect_equal(tz_lookup(pt, 3005, method = "accurate"), "Etc/GMT+9")
 })
+
+test_that("tz_lookup accurate method deals with overlapping timezones", {
+  coords <- c(85.18436, 41.11543) # centroid of https://en.wikipedia.org/wiki/Xinjiang_Time
+  expect_warning(tz <- tz_lookup_coords(coords[2], coords[1], method = "accurate"),
+                 "Some points are in areas with more than one timezone")
+  expect_equal(tz, "Asia/Shanghai; Asia/Urumqi")
+  coords_sfc <- sf::st_sfc(sf::st_point(coords), crs = 4326)
+  expect_warning(tz <- tz_lookup(coords_sfc, method = "accurate"),
+                 "Some points are in areas with more than one timezone")
+  expect_equal(tz, "Asia/Shanghai; Asia/Urumqi")
+  coords_sp <- as(coords_sfc, "Spatial")
+  expect_warning(tz <- tz_lookup(coords_sp, method = "accurate"),
+                 "Some points are in areas with more than one timezone")
+  expect_equal(tz, "Asia/Shanghai; Asia/Urumqi")
+})
diff --git a/tools/readme/unnamed-chunk-4-1.png b/tools/readme/unnamed-chunk-4-1.png
diff --git a/tools/readme/unnamed-chunk-5-1.png b/tools/readme/unnamed-chunk-5-1.png