Prepare CRAN Release 2.4.0 (#1336)

* Add missing `deprecate_warn()` to `src_sql()` * Minor cleanups * Organise NEWS * NEWS polish * Bump minimal R version * Bump dependencies * README * Specify timezone in tests * Add `db_col_types.default()` * Mention minor breaking change in `sql_not_supported()`
tidyverse · Aug 8, 2023 · 244ea87 · 244ea87
1 parent 5f92357
commit 244ea87
Show file tree

Hide file tree

Showing 6 changed files with 139 additions and 114 deletions.
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -17,39 +17,39 @@ License: MIT + file LICENSE
 URL: https://dbplyr.tidyverse.org/, https://github.com/tidyverse/dbplyr
 BugReports: https://github.com/tidyverse/dbplyr/issues
 Depends: 
-    R (>= 3.1)
+    R (>= 3.6)
 Imports: 
     blob (>= 1.2.0),
-    cli (>= 3.4.1),
-    DBI (>= 1.0.0),
-    dplyr (>= 1.1.0),
-    glue (>= 1.2.0),
+    cli (>= 3.6.1),
+    DBI (>= 1.1.3),
+    dplyr (>= 1.1.2),
+    glue (>= 1.6.2),
     lifecycle (>= 1.0.3),
     magrittr,
     methods,
-    pillar (>= 1.5.0),
+    pillar (>= 1.9.0),
     purrr (>= 1.0.1),
     R6 (>= 2.2.2),
-    rlang (>= 1.1.0),
-    tibble (>= 1.4.2),
+    rlang (>= 1.1.1),
+    tibble (>= 3.2.1),
     tidyr (>= 1.3.0),
     tidyselect (>= 1.2.0),
     utils,
-    vctrs (>= 0.5.2),
-    withr
+    vctrs (>= 0.6.3),
+    withr (>= 2.5.0)
 Suggests:
     bit64,
     covr,
     knitr,
     Lahman,
     nycflights13,
     odbc,
-    RMariaDB (>= 1.0.2),
+    RMariaDB (>= 1.2.2),
     rmarkdown,
-    RPostgres (>= 1.1.3),
+    RPostgres (>= 1.4.5),
     RPostgreSQL,
-    RSQLite (>= 2.2.15),
-    testthat (>= 3.0.2)
+    RSQLite (>= 2.3.1),
+    testthat (>= 3.1.10)
 VignetteBuilder: 
     knitr
 Config/Needs/website: tidyverse/tidytemplate

diff --git a/NEWS.md b/NEWS.md
@@ -1,157 +1,170 @@
 # dbplyr (development version)
 
+* Joins
+
+  * `*_join()` after `full_join()` works again (@mgirlich, #1178).
+
+  * `*_join()` now allows specifying the relationship argument. It must be
+    `NULL` or `"many-to-many"` (@bairdj, #1305).
+
+  * Queries now qualify `*` with the table alias for better compatibility (@mgirlich, #1003).
+
+  * `full_join()` can now handle column names that only differ in case (@ejneer, #1255).
+
+  * The `na_matches` argument of `semi_join()` and `anti_join()` works again
+    (@mgirlich, #1211).
+
+  * Joins now work again for Pool and Oracle connections (@mgirlich, #1177, #1181).
+
 * `dbplyr_pivot_wider_spec()` is now exported. Unlike `pivot_wider()` this can
   be lazy. Note that this will be removed soon after `pivot_wider_spec()`
   becomes a generic (@mgirlich).
 
-* Added translation for `str_detect()`, `str_starts()` and `str_ends()` with
-  fixed patterns (@mgirlich, #1009).
-
-* The `overwrite` argument of `db_copy_to()` now actually works.
-
-* `db_write_table()` and `db_save_query()` gain the `overwrite` argument.
+* `show_query()` and `remote_query()` gain the argument `sql_options` that allows
+  to control how the SQL is generated. It can be created via `sql_options()`
+  which has the following arguments:
+  * `cte`: use common table expressions?
+  * `use_star`: use `SELECT *` or explicitly select every column?
+  * `qualify_all_columns`: qualify all columns in a join or only the ambiguous ones?
+  (@mgirlich, #1146).
+
+* The `cte` argument of `show_query()` and `remote_query()` is deprecated (@mgirlich, #1146).
 
-* `distinct(.keep_all = TRUE)` now works for SQL Server (@mgirlich, #1053).
+## Breaking changes
 
-* `slice_min/max()` can now order by multiple variables like dplyr, e.g. use
-  `slice_min(lf, tibble(x, y))` (@mgirlich, #1167).
+* `remote_name()` now returns a string with the name of the table. To get the
+  qualified identifier use the newly added `remote_table()` (@mgirlich, #1280).
 
 * The functions `simulate_vars()` and `simulate_vars_is_typed()` were removed
   as they weren't used anymore and tidyselect now offers `tidyselect_data_proxy()`
   and `tidyselect_data_has_predicates()` (@mgirllich, #1199).
 
 * `translate_sql()` now requires the `con` argument (@mgirlich, #1311).
 
-* A `semi/anti_join()` where `y` is filtered is now inlined when possible (@mgirlich, #884).
+* `sql_query_append()`, `sql_query_insert()`, `sql_query_update()`,
+  `sql_query_upsert()`, and `sql_query_delete()` changed their arguments to
+  make them more consistent to the other `sql_query_*()` functions:
+
+  * `x_name` was renamed to `table`.
+  * `y` was renamed to `from` and must now be a table identifier or SQL instead
+    of a lazy table.
+  * `sql_query_append()` and `sql_query_insert()` have gained the argument `cols`.
 
-* `*_join()` now allows using specifying the relationship argument. It must be `NULL` or `"many-to-many"` (@bairdj, #1305).
+* The first argument of `ntile()` has been renamed from `order_by` to `x` to
+  match the interface of `dplyr::ntile()` (@mgirlich, #1242).
 
-* The columns generated when using a window function in `filter()` are now named
-  `col01` etc. instead of `q01()` (@mgirlich, #1258).
+* Removed argument `src` of `tbl_lazy()` after it has been deprecated for years
+  (@mgirlich, #1208).
 
-* `slice_*()` now supports the data masking pronouns `.env` and `.data` (@mgirlich, #1294).
+* Using `compute(temporary = FALSE)` without providing a name is now
+  deprecated (@mgirlich, #1154).
 
-* `tbl()` now informs when the user probably forgot to wrap the table identifier
-  with `in_schema()` or `sql()` (@mgirlich, #1287).
+* The `vars` argument of `translate_sql()` has been removed after it threw an
+  error for the last 7 years (@mgirlich).
 
-* Added `db_supports_table_alias_with_as()` to customise whether a backend supports
-  specifying a table alias with `AS` or not (@mgirlich).
+* `sql_not_supported()` now expects a function name without parentheses.
 
-* The translation of `between()` now also works for MS SQL when used in `mutate()`
-  (@mgirlich, #1241).
+## Improved SQL
 
-* MariaDB:
-  * `rows_update()` and `rows_patch()` now give an informative error when the
-    unsupported `returning` is used (@mgirlich, #1279).
-  * `rows_upsert()` now gives an informative error that it isn't supported
-    (@mgirlich, #1279).
-  * `rows_*()` use the column types of `x` when auto copying (@mgirlich, #1327).
+* A `semi/anti_join()` where `y` is filtered is now inlined when possible (@mgirlich, #884).
 
-* Oracle:
-  * Fix translation of `rows_upsert()` (@mgirlich, @TBlackmore, #1286)
-  * `head(n)` is now translated to `WHERE ROWNUM <= n` to also support old
-    versions <= 11.2 (@JeremyPasco, #1292).
+* A sequence of `union()` resp. `union_all()` now produces a flat query
+  instead of subqueries (@mgirlich, #1269).
 
+## Minor improvements and bug fixes
 * Teradata:
   * `as.Date(x)` is now translate to `CAST(x AS DATE)` again unless `x` is a
     string (@mgirlich, #1285).
 
 * Snowflake: 
   * `na.rm = TRUE` is now respected in `pmin()` and `pmax()` instead of being silently ignored (@fh-mthomson, #1329)
 
-* `remote_name()` now returns a string with the name of the table. To get the
-  qualified identifier use the newly added `remote_table()` (@mgirlich, #1280).
+* The `overwrite` argument of `db_copy_to()` now actually works.
 
-* Queries now qualify `*` with the table alias for better compatibility (@mgirlich, #1003).
+* `db_write_table()` and `db_save_query()` gain the `overwrite` argument.
 
-* Joins now work again for Pool and Oracle connections (@mgirlich, #1177, #1181).
+* The dots in `db_copy_to()` are now passed to `db_write_table()` (@mgirlich, #1237).
 
-* `show_query()` and `remote_query()` gain the argument `sql_options` that allows
-  to control how the SQL is generated. It can be created via `sql_options()`
-  which has the following arguments:
-  * `cte`: use common table expressions?
-  * `use_star`: use `SELECT *` or explicitly select every column?
-  * `qualify_all_columns`: qualify all columns in a join or only the ambiguous ones?
-  (@mgirlich, #1146).
-
-* The `cte` argument of `show_query()` and `remote_query()` is deprecated (@mgirlich, #1146).
+* Added translation for `str_detect()`, `str_starts()` and `str_ends()` with
+  fixed patterns (@mgirlich, #1009).
 
-* `any()` and `all()` now work for MS SQL (@ejneer, #1273).
+* Added translation for `nzchar()` (@MichaelChirico, @mgirlich, #1094).
 
-* Fixed negation of bit (boolean) fields in MS SQL (@ejneer, #1239)
+* Added translation for `runif()` (@mgirlich, #1200).
 
-* A sequence of `union()` resp. `union_all()` now produces a flat query
-  instead of subqueries (@mgirlich, #1269).
+* `sql_random()` is now deprecated. It was used to power `slice_sample()` which
+  is now done via the translation for `runif()` (@mgirlich, #1200).
+
+* `nth()`, `first()`, and `last()` now support the `na_rm` argument (@mgirlich, #1193).
 
-* Using `compute(temporary = FALSE)` without providing a name is now
-  deprecated (@mgirlich, #1154).
+* The translation of `if_any()` and `if_all()` is now wrapped in parentheses.
+  This makes sure it can be combined via `&` with other conditions (@mgirlich, #1153).
 
-* Added translation for `nzchar()` (@MichaelChirico, @mgirlich, #1094).
+* `slice_min/max()` can now order by multiple variables like dplyr, e.g. use
+  `slice_min(lf, tibble(x, y))` (@mgirlich, #1167).
 
-* MySQL/MariaDB:
-  * Fix translation of `as.numeric()`, `as.POSIXct()`, `as_datetime()`, and
-    `as.integer64()` (@avsdev-cw, #1189).
+* `slice_*()` now supports the data masking pronouns `.env` and `.data` (@mgirlich, #1294).
 
-* `full_join()` can now handle column names that only differ in case (@ejneer, #1255).
+* Can now use `select()` again after using `arrange(desc(x))` (@ejneer, #1240).
 
-* Subqueries now also get an alias for SQLite. This makes it consistent with
-  other backends and simplifies the implementation.
+* The columns generated when using a window function in `filter()` are now named
+  `col01` etc. instead of `q01()` (@mgirlich, #1258).
 
-* The translation of `if_any()` and `if_all()` is now wrapped in parentheses.
-  This makes sure it can be combined via `&` with other conditions (@mgirlich, #1153).
+* `tbl()` now informs when the user probably forgot to wrap the table identifier
+  with `in_schema()` or `sql()` (@mgirlich, #1287).
 
 * `pivot_wider()` now matches tidyr `NA` column handling (@ejneer #1238).
 
 * Using a function with a namespace in `across()` now works, e.g.
   `across(x, dplyr::dense_rank)` (@mgirlich, #1231).
 
-* The dots in `db_copy_to()` are now passed to `db_write_table()` (@mgirlich, #1237).
-
-* Can now use `select()` again after using `arrange(desc(x))` (@ejneer, #1240).
-
-* The first argument of `ntile()` has been renamed from `order_by` to `x` to
-  match the interface of `dplyr::ntile()` (@mgirlich, #1242).
-
-* Removed argument `src` of `tbl_lazy()` after it has been deprecated for years
-  (@mgirlich, #1208).
-
 * `sql_join_suffix()` gains the argument `suffix` so that methods can check
   whether the suffix is valid for the backend (@mgirlich).
 
-* `sql_query_append()`, `sql_query_insert()`, `sql_query_update()`,
-  `sql_query_upsert()`, and `sql_query_delete()` changed their arguments to
-  make them more consistent to the other `sql_query_*()` functions:
-
-  * `x_name` was renamed to `table`.
-  * `y` was renamed to `from` and must now be a table identifier or SQL instead
-    of a lazy table.
-  * `sql_query_append()` and `sql_query_insert()` have gained the argument `cols`.
-
-* The `na_matches` argument of `semi_join()` and `anti_join()` works again
-  (@mgirlich, #1211).
-
-* The `vars` argument of `translate_sql()` has been removed after it threw an
-  error for the last 7 years (@mgirlich).
-
-* Added translation for `runif()` (@mgirlich, #1200).
-
-* `sql_random()` is now deprecated. It was used to power `slice_sample()` which
-  is now done via the translation for `runif()` (@mgirlich, #1200).
+* Added `db_supports_table_alias_with_as()` to customise whether a backend supports
+  specifying a table alias with `AS` or not (@mgirlich).
 
-* DuckDB now supports the `returning` argument of `rows_*()`.
+## Backend specific improvements
 
-* `nth()`, `first()`, and `last()` now support the `na_rm` argument (@mgirlich, #1193).
+* Access
+  * Added translation for `!=` to `<>` (@erikvona, #1219).
 
-* `distinct()` + `head()` now work for Teradata (@mgirlich, #685).
+* DuckDB
+  * now supports the `returning` argument of `rows_*()`.
 
-* `copy_inline()` now works for MariaDB (@mgirlich, #1188).
+* MySQL/MariaDB:
+  * `rows_update()` and `rows_patch()` now give an informative error when the
+    unsupported `returning` argument is used (@mgirlich, #1279).
+  * `rows_upsert()` now gives an informative error that it isn't supported
+    (@mgirlich, #1279).
+  * `rows_*()` use the column types of `x` when auto copying `y` (@mgirlich, #1327).
+  * `copy_inline()` now works (@mgirlich, #1188).
+  * Fix translation of `as.numeric()`, `as.POSIXct()`, `as_datetime()`, and
+    `as.integer64()` (@avsdev-cw, #1189).
 
-* `*_join()` after `full_join()` works again (@mgirlich, #1178).
+* Oracle:
+  * Fix translation of `rows_upsert()` (@mgirlich, @TBlackmore, #1286)
+  * `head(n)` is now translated to `WHERE ROWNUM <= n` to also support old
+    versions <= 11.2 (@JeremyPasco, #1292).
 
-* The `rows_*()` functions now also work inside a transaction for Postgres
-  (@mgirlich, #1183).
+* Postgres
+  * The `rows_*()` functions now also work inside a transaction (@mgirlich, #1183).
 
-* Added translation for `!=` to `<>` for Microsoft Access (@erikvona, #1219).
+* SQLite
+  * Subqueries now also get an alias. This makes it consistent with other
+    backends and simplifies the implementation.
+
+* SQL Server
+  * `distinct(.keep_all = TRUE)` now works (@mgirlich, #1053).
+  * The translation of `between()` now also works when used in `mutate()`
+    (@mgirlich, #1241).
+  * `any()` and `all()` now work (@ejneer, #1273).
+  * Fixed negation of bit (boolean) fields (@ejneer, #1239)
+
+* Teradata
+  * `distinct()` + `head()` now work (@mgirlich, #685).
+  * `as.Date(x)` is now translate to `CAST(x AS DATE)` again unless `x` is a
+    string (@mgirlich, #1285).
 
 # dbplyr 2.3.3
 

diff --git a/R/db.R b/R/db.R
@@ -60,12 +60,14 @@ sql_join_suffix.DBIConnection <- function(con, suffix, ...) {
 db_sql_render <- function(con, sql, ..., cte = FALSE, sql_options = NULL) {
   check_bool(cte)
   if (cte) {
-    lifecycle::deprecate_soft(
+  lifecycle::deprecate_soft(
       when = "2.4.0",
       what = "db_sql_render(cte)",
       with = I("db_sql_render(sql_options = sql_options(cte = TRUE))")
     )
-    sql_options <- sql_options(cte = TRUE)
+    sql_options <- sql_options %||% sql_options(cte = TRUE)
+    out <- db_sql_render(con, sql, sql_options = sql_options)
+    return(out)
   }
 
   if (is.null(sql_options)) {
@@ -101,6 +103,12 @@ db_col_types.TestConnection <- function(con, table, call) {
 db_col_types.DBIConnection <- function(con, table, call) {
   NULL
 }
+# add a default method so that packages that haven't implemented `db_col_types()`
+# keep working, e.g. {Pool}
+#' @export
+db_col_types.default <- function(con, table, call) {
+  NULL
+}
 
 #' Options for generating SQL
 #'

diff --git a/R/src-sql.R b/R/src-sql.R
@@ -12,6 +12,11 @@
 #' @param con the connection object
 #' @param ... fields used by object
 src_sql <- function(subclass, con, ...) {
+  lifecycle::deprecate_warn(
+    when = "1.4.0",
+    what = "src_sql()",
+    always = TRUE
+  )
   subclass <- paste0("src_", subclass)
   structure(list(con = con, ...), class = c(subclass, "src_sql", "src"))
 }

diff --git a/R/verb-compute.R b/R/verb-compute.R
@@ -45,7 +45,6 @@ compute.tbl_sql <- function(x,
     if (!temporary) {
       lifecycle::deprecate_warn(
         "2.3.3",
-        # what = I(cli::format_inline("Not providing a name when {.code temporary = FALSE}"))
         what = "compute(name = 'must be provided when `temporary = FALSE`')"
       )
     }

diff --git a/README.md b/README.md
@@ -62,7 +62,7 @@ details). Printing it just retrieves the first few rows:
 mtcars2 <- tbl(con, "mtcars")
 mtcars2
 #> # Source:   table<mtcars> [?? x 11]
-#> # Database: sqlite 3.40.1 [:memory:]
+#> # Database: sqlite 3.41.2 [:memory:]
 #>      mpg   cyl  disp    hp  drat    wt  qsec    vs    am  gear  carb
 #>    <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
 #>  1  21       6  160    110  3.9   2.62  16.5     0     1     4     4
@@ -75,7 +75,7 @@ mtcars2
 #>  8  24.4     4  147.    62  3.69  3.19  20       1     0     4     2
 #>  9  22.8     4  141.    95  3.92  3.15  22.9     1     0     4     2
 #> 10  19.2     6  168.   123  3.92  3.44  18.3     1     0     4     4
-#> # … with more rows
+#> # ℹ more rows
 ```
 
 All dplyr calls are evaluated lazily, generating SQL that is only sent