Merge pull request #477 from cmu-delphi/lcb/slide-improvements-2024-06

Slide improvements pass

.Rbuildignore

@@ -16,3 +16,5 @@
 ^.lintr$
 ^DEVELOPMENT.md$
 man-roxygen
+^.venv$
+^sandbox.R$

.gitignore

@@ -13,3 +13,4 @@ docs
 renv/
 renv.lock
 .Rprofile
+sandbox.R

DESCRIPTION

@@ -1,10 +1,10 @@
 Type: Package
 Package: epiprocess
 Title: Tools for basic signal processing in epidemiology
-Version: 0.8.4
+Version: 0.9.0
 Authors@R: c(
     person("Jacob", "Bien", role = "ctb"),
-    person("Logan", "Brooks", email = "[email protected]", role = c("aut", "cre")),
+    person("Logan", "Brooks", , "[email protected]", role = c("aut", "cre")),
     person("Rafael", "Catoia", role = "ctb"),
     person("Nat", "DeFries", role = "ctb"),
     person("Daniel", "McDonald", role = "aut"),
@@ -15,23 +15,28 @@ Authors@R: c(
     person("Evan", "Ray", role = "aut"),
     person("Dmitry", "Shemetov", role = "ctb"),
     person("Ryan", "Tibshirani", role = "aut"),
-    person("Lionel", "Henry", role = "ctb", comment = "Author of included rlang fragments"),
-    person("Hadley", "Wickham", role = "ctb", comment = "Author of included rlang fragments"),
-    person("Posit", role = "cph", comment = "Copyright holder of included rlang fragments")
+    person("Lionel", "Henry", role = "ctb",
+           comment = "Author of included rlang fragments"),
+    person("Hadley", "Wickham", role = "ctb",
+           comment = "Author of included rlang fragments"),
+    person("Posit", role = "cph",
+           comment = "Copyright holder of included rlang fragments")
   )
-Description: This package introduces a common data structure for epidemiological
-    data reported by location and time, provides another data structure to
-    work with revisions to these data sets over time, and offers associated
-    utilities to perform basic signal processing tasks.
+Description: This package introduces a common data structure for
+    epidemiological data reported by location and time, provides another
+    data structure to work with revisions to these data sets over time,
+    and offers associated utilities to perform basic signal processing
+    tasks.
 License: MIT + file LICENSE
-Copyright: file inst/COPYRIGHTS
+URL: https://cmu-delphi.github.io/epiprocess/
+Depends:
+    R (>= 3.6)
 Imports:
     checkmate,
     cli,
     data.table,
     dplyr (>= 1.0.8),
     genlasso,
-    glue,
     ggplot2,
     glue,
     lifecycle (>= 1.0.1),
@@ -45,7 +50,8 @@ Imports:
     tidyselect (>= 1.2.0),
     tsibble,
     utils,
-    vctrs
+    vctrs,
+    waldo
 Suggests:
     covidcast,
     devtools,
@@ -59,22 +65,22 @@ VignetteBuilder:
     knitr
 Remotes:
     cmu-delphi/epidatr,
-    reconverse/outbreaks,
-    glmgen/genlasso
+    glmgen/genlasso,
+    reconverse/outbreaks
 Config/testthat/edition: 3
+Config/testthat/parallel: true
+Copyright: file inst/COPYRIGHTS
 Encoding: UTF-8
 LazyData: true
 Roxygen: list(markdown = TRUE)
 RoxygenNote: 7.3.2
-Depends:
-    R (>= 2.10)
-URL: https://cmu-delphi.github.io/epiprocess/
 Collate:
     'archive.R'
     'autoplot.R'
     'correlation.R'
     'data.R'
     'epi_df.R'
+    'epi_df_forbidden_methods.R'
     'epiprocess.R'
     'group_by_epi_df_methods.R'
     'methods-epi_archive.R'

NAMESPACE

@@ -2,8 +2,13 @@
 
 S3method("[",epi_df)
 S3method("names<-",epi_df)
+S3method(Summary,epi_df)
 S3method(arrange_canonical,default)
 S3method(arrange_canonical,epi_df)
+S3method(arrange_col_canonical,default)
+S3method(arrange_col_canonical,epi_df)
+S3method(arrange_row_canonical,default)
+S3method(arrange_row_canonical,epi_df)
 S3method(as_epi_df,data.frame)
 S3method(as_epi_df,epi_df)
 S3method(as_epi_df,tbl_df)
@@ -27,6 +32,7 @@ S3method(group_by,epi_df)
 S3method(group_by,grouped_epi_archive)
 S3method(group_by_drop_default,grouped_epi_archive)
 S3method(group_modify,epi_df)
+S3method(group_vars,grouped_epi_archive)
 S3method(groups,grouped_epi_archive)
 S3method(guess_period,Date)
 S3method(guess_period,POSIXt)
@@ -35,6 +41,7 @@ S3method(key_colnames,data.frame)
 S3method(key_colnames,default)
 S3method(key_colnames,epi_archive)
 S3method(key_colnames,epi_df)
+S3method(mean,epi_df)
 S3method(next_after,Date)
 S3method(next_after,integer)
 S3method(print,epi_archive)
@@ -71,6 +78,7 @@ export(filter)
 export(full_seq)
 export(geo_column_names)
 export(group_by)
+export(group_epi_df)
 export(group_modify)
 export(growth_rate)
 export(guess_period)
@@ -86,6 +94,7 @@ export(relocate)
 export(rename)
 export(revision_summary)
 export(slice)
+export(sum_groups_epi_df)
 export(time_column_names)
 export(ungroup)
 export(unnest)
@@ -145,6 +154,7 @@ importFrom(dplyr,everything)
 importFrom(dplyr,filter)
 importFrom(dplyr,group_by)
 importFrom(dplyr,group_by_drop_default)
+importFrom(dplyr,group_map)
 importFrom(dplyr,group_modify)
 importFrom(dplyr,group_vars)
 importFrom(dplyr,groups)

NEWS.md

@@ -4,11 +4,54 @@ Pre-1.0.0 numbering scheme: 0.x will indicate releases, while 0.x.y will indicat
 
 # epiprocess 0.9
 
+## Breaking changes
+
+- `epi_slide` interface has major breaking changes.
+  - All variables are now dot-prefixed to be more consistent with tidyverse
+    style for functions that allow tidyeval.
+  - The `before/after` arguments have been replaced with the `.window_size` and
+    `.align` arguments.
+  - `names_sep` has been removed. If you return data frames from your
+    computations:
+    - without a name, they will be unpacked into separate columns without name
+      prefixes
+    - with a name, it will become a packed data.frame-class column (see
+      `tidyr::pack`).
+  - `as_list_col` has been removed. You can now directly return a list from your
+    slide computations instead. If you were using `as_list_col=TRUE`, you will
+    need to wrap your output in a list.
+  - Ungrouped slides are no longer allowed in `epi_slide`. If you used this for
+    geographic aggregation up to national, consider using `sum_groups_epi_df`.
+  - Added `sum_groups_epi_df` to allow aggregation across key columns prior to
+    sliding.
+- `epix_slide` interface has major changes.
+  - All variables are now dot-prefixed to be more consistent with tidyverse
+    style for functions that allow tidyeval.
+  - `names_sep` has been removed. If you return data frames from your
+    computations:
+    - without a name, they will be unpacked into separate columns without name
+      prefixes
+    - with a name, it will become a packed data.frame-class column (see
+      `tidyr::pack`).
+  - `as_list_col` has been removed. You can now directly return a list from your
+    slide computations instead. If you were using `as_list_col=TRUE`, you will
+    need to wrap your output in a list.
+- `as_epi_df()` now checks that every group has unique time values and errors if
+  this is not the case. The same check is performed at the beginning of
+  `epi_slide()`. This check is currently not enforced in dplyr operations (like
+  for joins, mutates, or select), but we plan to add it in the future.
+- `as_epi_df()` or `as_epi_archive()` no longer accept `additional_metadata`.
+  Use the new `other_keys` arg to specify additional key columns, such as age
+  group columns or other demographic breakdowns. Miscellaneous metadata are no
+  longer handled by `epiprocess`, but you can use R's built-in `attr<-` instead
+  for a similar feature.
+
 ## Improvements
 
 - Added `complete.epi_df`, which fills in missing values in an `epi_df` with
   `NA`s. Uses `tidyr::complete` underneath and preserves `epi_df` metadata.
-- Inclusion of the function `revision_summary` to provide basic revision information for `epi_archive`s out of the box. (#492)
+- Inclusion of the function `revision_summary` to provide basic revision
+  information for `epi_archive`s out of the box. (#492)
 
 ## Bug fixes
 
@@ -21,6 +64,14 @@ Pre-1.0.0 numbering scheme: 0.x will indicate releases, while 0.x.y will indicat
 
 ## Breaking changes
 
+- `epi_df`'s are now more strict about what types they allow in the time column.
+  Namely, we are explicit about only supporting `Date` at the daily and weekly
+  cadence and generic integer types (for yearly cadence).
+- `epi_slide` `before` and `after` arguments are now require the user to
+  specific time units in certain cases. The `time_step` argument has been
+  removed.
+- `epix_slide` `before` argument now defaults to `Inf`, and requires the user to
+  specify units in some cases. The `time_step` argument has been removed.
 - `detect_outlr_stl(seasonal_period = NULL)` is no longer accepted. Use
   `detect_outlr_stl(seasonal_period = <value>, seasonal_as_residual = TRUE)`
   instead. See `?detect_outlr_stl` for more details.
@@ -63,6 +114,12 @@ Pre-1.0.0 numbering scheme: 0.x will indicate releases, while 0.x.y will indicat
   are similar functions for `geo` and `version`).
 - Fixed bug where `epix_slide_ref_time_values_default()` on datetimes would
   output a huge number of `ref_time_values` spaced apart by mere seconds.
+- In `epi_slide()` and `epix_slide()`:
+  - Multiple "data-masking" tidy evaluation expressions can be passed in via
+    `...`, rather than just one.
+  - Additional tidy evaluation features from `dplyr::mutate` are supported: `!!
+name_var := value`, unnamed expressions evaluating to data frames, and `=
+NULL`; see `?epi_slide` for more details.
 
 ## Cleanup
 
@@ -71,17 +128,6 @@ Pre-1.0.0 numbering scheme: 0.x will indicate releases, while 0.x.y will indicat
   of `epi_df`s to let `{epipredict}` work more easily with other libraries (#471).
 - Removed some external package dependencies.
 
-## Breaking Changes
-
-- `epi_df`'s are now more strict about what types they allow in the time column.
-  Namely, we are explicit about only supporting `Date` at the daily and weekly
-  cadence and generic integer types (for yearly cadence).
-- `epi_slide` `before` and `after` arguments are now require the user to
-  specific time units in certain cases. The `time_step` argument has been
-  removed.
-- `epix_slide` `before` argument now defaults to `Inf`, and requires the user to
-  specify units in some cases. The `time_step` argument has been removed.
-
 # epiprocess 0.7.0
 
 ## Breaking changes: