Fix warnings&errors in examples&tests, keep archive_cases_dv in sync

- Fix unmuffled warnings and errors in the tests and examples
  - Introduce the S3 class `"epiprocess__snapshot_as_of_last_update_version"` to
    selectively muffle this type of warning.
  - Fix some things that only arise when running checks or examples from a fresh
    R session / one that does not have epiprocess, dplyr, and tibble loaded.
- When re-building `archive_cases_dv`, use `compactify=FALSE` to not mess up the
  current tests.
- Don't store `archive_cases_dv` directly, as its R6 implementation will be
  stored with it; instead, just store its DT and reconstruct the archive object
  when requested (use a promise and mark it to @export).

Author: lcbrooks

NAMESPACE

@@ -20,6 +20,7 @@ S3method(tail,epi_df)
 S3method(ungroup,epi_df)
 S3method(unnest,epi_df)
 export("%>%")
+export(archive_cases_dv)
 export(arrange)
 export(as_epi_archive)
 export(as_epi_df)

R/archive.R

@@ -196,7 +196,7 @@ epi_archive =
               DT = rm_locf(DT)
             } else {
               # Create empty data frame for nrow(elim) to be 0
-              elim = tibble()
+              elim = tibble::tibble()
             }
 
             # Warns about redundant rows
@@ -280,7 +280,7 @@ epi_archive =
               Abort("`max_version` must be at most `max(DT$max_version)`.")
             }
             if (max_version == self_max) {
-              Warn("Getting data as of the latest version possible. For a variety of reasons, it is possible that we only have a preliminary picture of this version (e.g., the upstream source has updated it but we have not seen it due to latency in synchronization). Thus, the snapshot that we produce here might not be reproducible at a later time (e.g., when the archive has caught up in terms of synchronization).")
+              Warn("Getting data as of the latest version possible. For a variety of reasons, it is possible that we only have a preliminary picture of this version (e.g., the upstream source has updated it but we have not seen it due to latency in synchronization). Thus, the snapshot that we produce here might not be reproducible at a later time (e.g., when the archive has caught up in terms of synchronization).", class="epiprocess__snapshot_as_of_last_update_version")
             }
 
             # Filter by version and return
@@ -525,7 +525,7 @@ epi_archive =
 #'
 #' @export
 #' @examples 
-#' df <- data.frame (geo_value  = c(replicate(2, "ca"), replicate(2, "fl")),
+#' df <- data.frame (geo_value  = c(rep("ca", 2), rep("fl", 2)),
 #'                  county = c(1, 3, 2, 5),
 #'                  time_value = c("2020-06-01", 
 #'                  "2020-06-02", 

R/data.R

@@ -41,5 +41,10 @@
 #' 
 #' COVID-19 Data Repository by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University. 
 #' \url{https://github.com/CSSEGISandData/COVID-19}
+#'
+#' @export
 "archive_cases_dv"
 
+# Like normal data objects, set archive_cases_dv up as a promise, so it doesn't
+# take unnecessary space. This also avoids a need for @include tags.
+delayedAssign("archive_cases_dv", as_epi_archive(archive_cases_dv_dt, compactify=FALSE))

R/methods-epi_archive.R

@@ -27,9 +27,25 @@
 #'   ```
 #' 
 #' @export 
-#' @examples 
-#' epix_as_of(x = archive_cases_dv, 
-#'            max_version = max(archive_cases_dv$DT$version))
+#' @examples
+#'
+#' range(archive_cases_dv$DT$version) # 2020-06-02 -- 2020-06-15
+#'
+#' epix_as_of(x = archive_cases_dv,
+#'            max_version = as.Date("2020-06-12"))
+#'
+#' # When fetching a snapshot as of the latest version with update data in the
+#' # archive, a warning is issued as this update data might not yet be finalized
+#' # (for example, if data versions are labeled with dates, these versions might be
+#' # overwritten throughout the day if the data can be updated multiple times per
+#' # day; when we build an archive based on special update-data queries all made at
+#' # the same time, the latest available update might still be subject to change,
+#' # but previous versions should be finalized). We can muffle such warnings with
+#' # the following pattern:
+#' withCallingHandlers({
+#'   epix_as_of(x = archive_cases_dv,
+#'              max_version = max(archive_cases_dv$DT$version))
+#' }, epiprocess__snapshot_as_of_last_update_version = function(wrn) invokeRestart("muffleWarning"))
 epix_as_of = function(x, max_version, min_time_value = -Inf) {
   if (!inherits(x, "epi_archive")) Abort("`x` must be of class `epi_archive`.")
   return(x$as_of(max_version, min_time_value))
@@ -74,10 +90,10 @@ epix_as_of = function(x, max_version, min_time_value = -Inf) {
 #' # create two example epi_archive datasets
 #' x <- archive_cases_dv$DT %>% 
 #'   dplyr::select(geo_value,time_value,version,case_rate) %>% 
-#'   as_epi_archive()
+#'   as_epi_archive(compactify=TRUE)
 #' y <- archive_cases_dv$DT %>% 
 #'   dplyr::select(geo_value,time_value,version,percent_cli) %>% 
-#'   as_epi_archive()
+#'   as_epi_archive(compactify=TRUE)
 #'   
 #' # a full join stored in x
 #' epix_merge(x, y, all = TRUE) 

R/sysdata.rda

@@ -8,20 +8,20 @@ archive_cases_dv <- covidcast(
   signals = "smoothed_adj_cli",
   time_type = "day",
   geo_type = "state",
-  time_value = epirange(20200601, 20200615),
+  time_values = epirange(20200601, 20200615),
   geo_values = "ca,fl",
   issues = epirange(20200601, 20200615)
 ) %>% 
   fetch_tbl() %>%
   select(geo_value, time_value, version = issue, percent_cli = value) %>%
-  as_epi_archive(compactify=TRUE)
+  as_epi_archive(compactify=FALSE)
 
 case_rate <- covidcast(
   data_source = "jhu-csse",
   signals = "confirmed_7dav_incidence_prop",
   time_type = "day",
   geo_type = "state",
-  time_value = epirange(20200601, 20200615),
+  time_values = epirange(20200601, 20200615),
   geo_values = "ca,fl",
   issues = epirange(20200601, 20200615)
 ) %>%
@@ -31,4 +31,11 @@ case_rate <- covidcast(
 
 epix_merge(archive_cases_dv, case_rate, all = TRUE)
 
-usethis::use_data(archive_cases_dv, overwrite = TRUE)
+# If we directly store an epi_archive R6 object as data, it will store its class
+# implementation there as well. To prevent mismatches between these stored
+# implementations and the latest class definition, don't store them as R6
+# objects; store the DT and construct the R6 object on request.
+
+archive_cases_dv_dt = archive_cases_dv$DT
+
+usethis::use_data(archive_cases_dv_dt, overwrite = TRUE, internal=TRUE)

data-raw/archive_cases_dv.R

@@ -48,7 +48,7 @@ dt <- row_replace(74,73,74) # Not LOCF
 
 dt_true <- as_tibble(as_epi_archive(dt,compactify=TRUE)$DT)
 dt_false <- as_tibble(as_epi_archive(dt,compactify=FALSE)$DT)
-dt_null <- as_tibble(as_epi_archive(dt,compactify=NULL)$DT)
+dt_null <- suppressWarnings(as_tibble(as_epi_archive(dt,compactify=NULL)$DT))
 
 test_that("Warning for LOCF with compactify as NULL", {
   expect_warning(as_epi_archive(dt,compactify=NULL))
@@ -73,12 +73,14 @@ test_that("as_of utilizes LOCF even after removal of LOCF values",{
   ea_true <- as_epi_archive(dt,compactify=TRUE)
   ea_false <- as_epi_archive(dt,compactify=FALSE)
 
-  epix_as_of(ea_true,max(ea_true$DT$version))
-  
   # Row 22, an LOCF row corresponding to the latest version, but for the
   # date 2020-06-02, is omitted in ea_true
-  as_of_true  <- ea_true$as_of(max(ea_true$DT$version))
-  as_of_false <- ea_false$as_of(max(ea_false$DT$version))
+  expect_warning({
+    as_of_true  <- ea_true$as_of(max(ea_true$DT$version))
+  }, class = "epiprocess__snapshot_as_of_last_update_version")
+  expect_warning({
+    as_of_false <- ea_false$as_of(max(ea_false$DT$version))
+  }, class = "epiprocess__snapshot_as_of_last_update_version")
 
   expect_identical(as_of_true,as_of_false)
 })