• Explicit permissions (#2254).

• Use styler from main branch (#2252).

• Need to install R on Ubuntu 24.04 (#2250).

• Use Ubuntu 24.04 and styler PR (#2248).

• Use styler from main branch (#2252).

• Need to install R on Ubuntu 24.04 (#2250).

• Use Ubuntu 24.04 and styler PR (#2248).

• Need to install R on Ubuntu 24.04 (#2250).

• Use Ubuntu 24.04 and styler PR (#2248).

• Add support for Redshift connections (@owenjonesuob, #2215).

• Drop crayon and mockr dependencies (@olivroy, #2220).

• Use index.md.

• Tweak formatting (@salim-b, #2232).

• Fix typo (@salim-b, #2218).

• Fix compatibility with waldo >= 0.6.0 (#2240).

• Establish compatibility with igraph >= 2.0.0 (#2187) and withr 3.0.0 (#2184).

• Reexport tibble::glimpse() instead of pillar::glimpse() to avoid pillar dependency with roxygen2 7.3.0 (#2179).

• dm_sql() now processes table_names with dbplyr::escape(), therefore also accepting dbplyr objects (#2129).

• Adapt to DBI >= 1.2.0 (#2148, #2155).

• copy_dm_to() creates string columns of necessary lengths for MariaDB and SQL Server. This worked before for SQL Server in dm 1.0.5, now also works on MariaDB (#311, #2066, #2082).

• Explicitly fail on compute(temporary = TRUE), which never worked correctly (#2059, #2103).

• Warn about DuckDB not supporting autoincrementing primary keys (#2099).

• Make check_suggested() a standalone (#2054).

• Tweak vignette for compute(temporary = TRUE).

• Update documentation of check_suggested() (@olivroy, #2055).

• Speed up dm(), new_dm(), as_dm() and dm_validate(). dm() and as_dm() no longer call dm_validate() (#2108).

• Add test for copy_dm_to(table_names = ) (#250, #2101).

• Work around test failures for dbplyr 2.4.0.

• Remove most skips from tests (#2052).

• copy_dm_to() now warns unconditionally on unsupported arguments, and fails if copy_to is provided (#1944). Use the new dm_sql() function as a replacement for copy_dm_to(copy_to = ) (#1915, #2011, @jangorecki).

• New json_unnest() and json_unpack(), currently implemented for data frames only (#991, #997).

• dm_rows_append() also works for local dm, with support for autoincrement primary keys (#1727, #1745).

• Breaking change: Add check_dots_empty() calls (#1929, #1943).

• Test MySQL on GHA (#1940).

• Improve MySQL compatibility regarding learning of database schemas and checking of constraints (#1938).

• Breaking change: Add check_dots_empty() calls (#1929, #1943).

• Compatibility with duckdb 0.9.1.

• Minor fixes in dm_pack_tbl() and dm_unwrap_tbl() (#1947).

• Use rlang::check_installed() internally to install missing suggested packages on the fly (@olivroy, #1348, #2036, #2039, #2040).

• Use vectorized rlang::is_installed()to decide if examples should be run (@olivroy, #2043).

• Recategorize and describe function reference.

• Better error and information messages when querying keys.

• collect.zoomed_dm() shows a more helpful error message (#1929, #1945).

• Add information on default font size to ?dm_draw (#1935).

• Add db-* rules to Makefile to simplify Docker-based database setup.

• Remove curly braces, add \pkg (@olivroy, #2042).

• Replace tibble() by fast_tibble() (@mgirlich, #1928).

• Replace superseded dplyr::recode() (@mgirlich, #1927).

• Remove most skips from tests (#2052).

• Add explicit unique key to dm_for_filter().

• Add Postgres test for dm_sql().

• Switch internal testing to MariaDB.

• Fast offline checks with new "DM_OFFLINE" environment variable.

• New GHA checks for the case of missing suggested packages (#1952).

• Make check_suggested() a standalone (#2054).

• Backport changes from attempted CRAN release (#2046).

• Move magrittr (#1975, #1983), DBI (#1974), and pillar (#1976) to "Suggests".

• Require RMariaDB 1.3.0, work around tidyverse/dbplyr#1190 and tidyverse/dbplyr#1195 (#1989).

• Prefer map*() over lapply() and vapply().

• styler::style_pkg(scope = "tokens").

• Rename internal new_dm3() to dm_from_def() (#1225, #1949).

• Remove dead code (#979, #1950, #1871).

• Reorganize build_copy_queries() (#1923).

• Avoid dbplyr::ident_q() (#1788).

• Add ellipsis to tbl_sum() signature (#1941).

• Compare version returned by getRversion() with string instead of number.

• Work around vctrs bug in jsonlite 1.8.5.

• dm_from_con() gains .names argument for pattern-based construction of table names in the dm object (@owenjonesuob, #1790).

• New dm_set_table_descriptions(), dm_get_table_descriptions() and dm_reset_table_descriptions() to set table labels as persistent attributes of the table object (#1888).

• dm_from_con() can retrieve multiple schemas, pass a character vector to the schema argument (@owenjonesuob, #1533, #1789).

• build_copy_queries() and db_learn_from_db() improvements (@samssann, #1642, #1677, #1739).

• UK support for dm_draw() (#1731, #1877).

• Allow for additional description of tables in dm_draw() (#1875, #1876).

• Establish compatibility with dbplyr 2.3.3 and 2.4.0 (@mgirlich, #1919).

• In copy_dm_to(), call collect() only when copying data, table by table (@jangorecki, #1900).

• Use roxyglobals (#1838).

• Require purrr >= 1.0.0 for list_c() (#1847, #1848).

• Add table description to diagram in README.

• Tweak testing instructions. Mention Makefile in CONTRIBUTING.md. Describe Docker setup (#1898).

• Vignette corrections (@MikeJohnPage, #1882).

• Avoid tidyverse package.

• Fix local tests (#1921).

• Progress bars for dm_wrap_tbl() and dm_unwrap_tbl() (#835, #1450).

• Add cheat sheet as a vignette (#1653).

• Suggest creating a function for your database dm object (#1827, #1828).

• Add alternative text to author images for pkgdown website (#1804).

• Compatibility with dev jsonlite (#1837).

• Remove tidyverse dependency (#1798, #1834).

• Minimal patch to fix multiple match updates (@DavisVaughan, #1806).

• Adapt to rlang 1.1.0 changes (#1817).

• Make sure {dm} passes "noSuggests" workflow (#1659).

• dm_add_pk() gains autoincrement argument (#1689), autoincrement primary keys are configured on the database with copy_dm_to() (#1696).

• New dm_add_uk(), dm_rm_uk() and dm_get_all_uks() functions for explicit support of unique keys (#622, #1716).

• dm_get_all_pks() and dm_get_all_fks() return output in the order of table or parent_table argument (#1707).

• Improve error message for dm_add_pk() when the columns argument is missing (#1644, #1646).

• dm_get_all_pks(), dm_get_all_fks(), and dm_get_all_uks() require unquoted table names as input, for consistency with other parts of the API (#1741).

• dm_examine_constraints() works for dm objects on the database with compound keys (#1713).

• Update pkgdown URL to https://dm.cynkra.com/ (#1652).

• Fix link rot (#1671).

• Require dplyr >= 1.1.0 and lifecycle >= 1.0.3 (#1771, #1637).

• Checks pass if all suggested packages are missing (#1659).

• Fix r-devel builds (#1776).

• dm_unpack_tbl() sets PK before FK (#1715).

• Clean up dm_rows_append() implementation (#1714).

• dm() accepts tables that are of class "tbl_sql" but not "tbl_dbi" (#1695, #1710).

• Use correctly typed missing value for lists (@DavisVaughan, #1686).

• Avoid running example without database connection.

• dm_from_con() can use multiple schemata (@mgirlich, #1441, #1449).

• pack_join(keep = TRUE) preserves order of packed columns (#1513, #1514).

• pack_join(keep = TRUE) keeps keys of y in the resulting packed column (#1451, #1452).

• New json_pack.tbl_lazy() and json_nest.tbl_lazy() (#969, #975).

• dm_paste() gives correct output for factor columns with many levels (#1510, #1511).

• Fix compatibility with duckdb 0.5.0 (#1509, #1518).

• Refactor dm_unwrap_tbl() so it builds a "unwrap plan" first (#1446, #1447).

• Reenable dm_rows_update() test (#1437).

• New dm_deconstruct() creates code to deconstruct a dm object into individual keyed tables via pull_tbl(keyed = TRUE) (#1354).

• Use dm_ptype() in dm_gui(), generate better code (#1353).

• New dm_gui() for interactive editing of dm objects (#1076, #1319).

• dm_get_tables() and pull_tbl() gain a new keyed = FALSE argument. If set to TRUE, table objects of class "dm_keyed_tbl" are returned. These objects inherit from the underlying data structure (tibble or lazy table), keep track of primary and foreign keys, and can be used later on in a call to dm() to recreate a dm object with the keys (#1187).

• New by_position argument to check_subset(), check_set_equality(), check_cardinality_...() and examine_cardinality() (#1253).

• dm() accepts dm objects (#1226).

• dm_examine_constraints() honors implicit unique keys defined by foreign keys (#1131, #1209).

• dm_filter() is now stable, with a new API that avoids exposing an intermediate state with filters not yet applied, with a compatibility wrapper (#424, #426, #1236).

• check_cardinality_...(), examine_cardinality(), check_subset() and check_set_equality() are now stable and consistently use a common interface with arguments named x, y, x_select and y_select, with compatibility wrappers (#1194, #1229).

• dm_examine_cardinalities() and dm_examine_constraints() are now stable with a new signature and a compatibility wrapper (#1193, #1195).

• dm_apply_filters(), dm_apply_filters_to_tbl() and dm_get_filters() are deprecated (#424, #426, #1236).

• dm_disambiguate_cols() adds table names as a suffix by default, and gains a .position argument to restore the original behavior. Arguments sep and quiet are renamed to .sep and .quiet (#1293, #1327).

• dm_squash_to_tbl() is deprecated in favor of the new .recursive argument to dm_flatten_to_tbl(). Arguments start and join are renamed to .start and .join (#1272, #1324).

• dm_rm_tbl() is deprecated in favor of dm_select_tbl() (#1275).

• dm_bind() and dm_add_tbl() are deprecated in favor of dm() (#1226).

• rows_truncate() and dm_rows_truncate() are deprecated, because they use DDL as opposed to all other verbs that use DML (#1031, #1321).

• All internal S3 classes now use the "dm_" prefix (#1285, #1339).

• Add ellipses to all generics (#1298).

• Reexport tibble() (#1279).

• dm_ptype(), dm_financial() and dm_pixarfilms() are stable now (#1254).

• Turn all "questioning" functions to "experimental" (#1030, #1237).

• is_unique_key()uses vctrs::vec_count() on local data frames for speed (@eutwt, #1247).

• check_key() uses vctrs::vec_duplicate_any() on local data frames for speed (@eutwt, #1234).

• dm_draw() works if a table name has a space (#1219).

• Don't print rule in glimpse.dm() for empty dm() (#1208).

• Work around ANSI escape issues in CRAN rendering of vignette (#1156, #1330).

• Fix column names in ?dm_get_all_pks (#1245).

• Improve contrast for display of dm_financial() (#1073, #1250).

• Add contributing guide (#1222).

• Use sensible node and edge IDs, corresponding to the data model, in SVG graph (#1214).

• Tests for datamodelr code (#1215).

• Implement glimpse() for zoomed_df (@IndrajeetPatil, #1003, #1161).

• Remove message about automated key selection with the select argument in joins on zoomed_df (@IndrajeetPatil, #1113, #1176).

• dm_from_con(learn_keys = TRUE) works for MariaDB (#1106, #1123, #1169, @maelle), and for compound keys in Postgres (#342, #1006, #1016) and SQL Server (#342).

• New json_pack_join(), json_nest_join(), json_pack() and json_nest(), similar to pack_join(), dplyr::nest_join(), tidyr::pack() and tidyr::nest(), but create character columns (#917, #918, #973, #974).

• nest_join() and pack_join() support zoomed_df objects (#1119, @IndrajeetPatil).

• Marked stable functions as stable, in particular dm() and related functions (#1032, #1040).

• Remove own rows_*() implementation for lazy tables, they are now available in dbplyr >= 2.2.0 (#912, #1024, #1028).

• Deprecate dm_join_to_tbl(), dm_is_referenced() and dm_get_referencing_tables() (#1038).

• New dm_validate() replaces now deprecated validate_dm() (#1033).

• dm_get_con() and dm_get_filters() use dm as argument name (#1034, #1036).

• Mark ... in dm_flatten_to_tbl() as experimental (#1037).

• Add ellipses to dm_disambiguate_cols(), dm_draw(), dm_examine_constraints(), dm_nycflights13() and dm_pixarfilms() (#1035).

• New dm_from_con(), soft-deprecated dm_from_src() (#1014, #1018, #1044).

• Moved pack_join() arguments past the ellipsis for consistency (#920, #921).

• Compatibility fix for writing to SQL Server tables with dbplyr >= 2.2.0.

• The pkgdown site now uses BS5 for greater readability (#1067, @maelle).

• Better message for dm_rows_...() functions if the in_place argument is missing (@IndrajeetPatil, #414, #1160).

• Better message for learning error (#1081).

• Greatly improved consistency, content, and language across all articles (@IndrajeetPatil, #1056, #1132, #1157, #1166, #1079, #1082, #1098, #1100, #1101, #1103, #1112, #1120, #1158, #1175).

• Tweaks of intro vignette and README (#1066, #1075, @maelle).

• Document glimpse() S3 method for dm (@IndrajeetPatil, #1121).

• Update credentials to fallback databases for dm_financial() hosted on pacha.dev (#916, @pachadotdev), also used now for vignettes (#1118) and in dm_from_con() example (#993).

• Update license year (#1029).

• Switch to duckdb as default database backend (#1179).

• Test duckdb and MariaDB on GHA (#1091, #1136).

• pack_join() works correctly if name is the same as an existing column in either table. In some cases a column is overwritten, this is consistent with nest_join() behavior (#864, #865).
• Messages that suggest the installation of optional packages are shown only once per session (#852).
• dm_insert_zoomed() uses the color from the zoomed table for the new table (#750, #863).
• copy_dm_to() consumes less memory and is faster when writing to SQL Server (#855).

• Remove extra spaces in output when examining constraints with compound keys (#868).
• Fix column tracking for foreign keys (#856, #857).
• copy_dm_to() shows progress bars again (#850, #855).
• Progress bars use the console width (#853).
• Avoid calling dbAppendTable() for zero-row tables (#847).

• Require rlang 1.0.1 (#840).

• New dm_wrap_tbl(), dm_unwrap_tbl(), dm_nest_tbl(), dm_unnest_tbl(), dm_pack_tbl() and dm_unpack_tbl() (#595, #733, #737).
• New dm_examine_cardinality() (#264, #735).
• New pack_join() generic and method for data frames, the same to tidyr::pack() as dplyr::nest_join() is to tidyr::nest() (#721, #722).
• dm_pixarfilms() is exported and gains a consistent = FALSE argument; if TRUE the data is modified so that all referential constraints are satisfied (#703, #707, #708, @erictleung).

• db_schema_...() functions no longer pro-actively check for schema existence (#672, #815, #771).
• db_schema_list.Microsoft SQL Server no longer ignoring schemas for which the owner cannot be found (#815, #771).
• copy_dm_to() works with DuckDB again, the SQL statements to define the tables and indexes are now created by us (#701, #709).

• Establish compatibility with rlang 1.0.0 (#756).
• Simplify database checks on GitHub Actions (#712).

• New dm_pixarfilms() creates a dm object with data from the {pixarfilms} package (#600, @erictleung).
• check_cardinality_0_1(), check_cardinality_0_n(), check_cardinality_1_1(), check_cardinality_1_n(), and examine_cardinality() now support compound keys (#524).
• check_subset() and check_set_equality() support compound keys (#523).
• dm_paste() adds the on_delete argument to dm_add_fk() (#673).
• dm_disambiguate_cols() also disambiguates columns used in keys, to support correct disambiguation for compound keys (#662).
• dm_disambiguate_cols() now emits the source code equivalent of a renaming operation (#684).
• dm_examine_constraints() uses backticks to surround table names (#687).

• decompose_table() now avoids creating NA values in the key column (#580).
• dm_draw() works with empty tables (#585).

• Fix compatibility with dplyr 1.0.8 (#698).

• mutate(), transmute(), distinct() and summarize() now support dplyr::across() and extra arguments (#640).
• Key tracking for the first three verbs is less strict and based on name equality (#663).
• relocate() now works on zoomed dm objects (#666).
• dm_add_fk() gains on_delete argument which copy_dm_to() picks up and translates to an ON DELETE CASCADE or ON DELETE NO ACTION specification for the foreign key (#649).
• dm_copy_to() defines foreign keys during table creation, for all databases except DuckDB. Tables are created in topological order (#658). For cyclic relationship graphs, table creation is attempted in the original order and may fail (#664).
• waldo::compare() shows better output for dm objects (#642).
• dm_paste() output uses trailing commas in the dm::dm() and tibble::tibble() calls, and sorts column attributes by name, for better modularity (#641).

• New db_schema_create(), db_schema_drop(), db_schema_exists() and db_schema_list() replace the corresponding sql_schema_*() functions, the latter are soft-deprecated (#670). The connection argument to db_schema_*() is called con, not dest (#668).

• copy_dm_to() and sql_create_schema() no longer actively check for schema existence (#644, #660).
• Add newline after OUTPUT clause for SQL Server (#647).
• Fix sql_rows_delete() with returning argument for SQL Server (#645).

• Remove method only needed for RSQLite < 2.2.8, add warning if loaded RSQLite version is <= 2.2.8 (#632).
• Adapt MSSQL tests to testthat update (#648).

• rows_insert(), rows_update() and rows_delete() gain returning argument. In combination with in_place = TRUE this argument makes the newly inserted rows accessible via get_returning_rows() after the operation completes (#593, @mgirlich).
• Implement rows_patch() for DBI connections (#610, @mgirlich).
• Use NO ACTION instead of CASCADE in foreign key constraints to permit self-references.
• dm_from_src() supports pool::Pool objects (#599, @moodymudskipper).
• Better error message for dm_rows_update() and related functions for dm objects with tables without primary key (#592).
• glimpse() is implemented for dm objects (#605).
• Support DuckDB in rows_insert(), rows_update() and rows_delete() (#617, @mgirlich).

• Fix dm_zoom_to() for dm objects with an empty table (#626, @moodymudskipper).
• Avoid generating invalid dm objects in some corner cases (#596).

• sql_schema_list() supports pool::Pool objects (#633, @brancengregory).
• Establish compatibility with pillar 1.6.2, vctrs > 0.3.8 and rlang > 0.4.11 (#613).
• Use check_suggested() everywhere (#572, @moodymudskipper).
• Add CI run for validating all new dm objects (#597).

• Fix rows_truncate() in interactive mode (#588).

• Implement rows_delete() for databases (#589).

• Skip examples that might require internet access on non-CI platforms.

• dm_examine_constraints(), dm_rows_insert() and related, copy_dm_to() and collect.dm() show progress bars in interactive mode via the progress package. The new progress = NA argument controls the behavior (#262, @moodymudskipper).
• copy_dm_to() gains a copy_to argument to support other ways of copying data to the database (#582).

• Always run database tests on sqlite for df source.
• Establish compatibility with testthat > 3.0.2 (#566, @moodymudskipper).

• dm_get_all_fks() returns a data frame with a parent_key_cols instead of a parent_pk_cols column (introduced in dm 0.2.0), to reflect the fact that a foreign key no longer necessarily points to a primary key (#562).
• *_pk() and *_fk() functions now verify that the dots are actually empty (#536).
• dm_get_pk() is deprecated in favor of dm_get_all_pks() (#561).
• dm_has_fk() and dm_get_fk() are deprecated in favor of dm_get_all_fks() (#561).

• dm_add_fk() gains ref_columns argument that supports creating foreign keys to non-primary keys (#402).
• dm_get_all_pks() gains table argument for filtering the returned primary keys (#560).
• dm_get_all_fks() gains parent_table argument for filtering the returned foreign keys (#560).
• dm_rm_fk() gains an optional ref_columns argument. This function now supports removal of multiple foreign keys filtered by parent or child table or columns, with a message (#559).
• dm_rm_pk() gains columns argument and allows filtering by columns and by tables or removing all primary keys. The rm_referencing_fks argument has been deprecated in favor of the new fail_fk argument (#558).
• dm_get_all_fks() has been optimized for speed and no longer sorts the keys (#560).
• dm operations are now slightly faster overall.

• The internal data structure for a dm object has changed to accommodate foreign keys to other columns than the primary key. An upgrade message is shown when working with a dm object from an earlier version, e.g. if it was loaded from a cache or an .rds file (#402).
• Drop "dm_v1" class from dm objects again, this would have made every S3 dispatch more costly. Relying on an internal "version" attribute instead (#547).

• Deprecate dm_get_src() tbl.dm(), src_tbls.dm(), copy_to.dm(). These functions have better alternatives and use the notion of a "data source" which is being phased out of dplyr (#527).
• *_pk() and *_fk() functions gain an ellipsis argument that comes before check, force and rm_referencing_fks arguments (#520).

• dm_add_pk() and dm_add_fk() support compound keys via the c() notation, e.g. dm_add_pk(dm, table, c(col1, col2)). dm_nycflights13() returns a data model with compound keys by default. Use compound = FALSE to return the data model from dm v0.1.13 or earlier (#3).
• dm_get_all_fks() includes parent_pk_cols column that describes the primary key columns of the parent table (#335).
• dm_from_src() supports the schema argument also for MariaDB and MySQL databases (#516).
• dm objects now inherit from "dm_v1" in addition to "dm", to allow backward-compatible changes of the internal format (#521).
• Use hack to create compound primary keys on the database (#522).
• dm_examine_constraints() and other check functions count the number of rows that violate constraints for primary and foreign keys (#335).
• copy_dm_to(set_key_constraints = FALSE) downgrades unique indexes to regular indexes (#335).
• rows_truncate() implemented for data frames (#335).
• dm_enum_fk_candidates() enumerates column in the order they apper in the table (#335).

• dm_draw() gains column_types argument, if TRUE the column type is shown for each displayed column (#444, @samssann).
• copy_dm_to() gains schema argument (#432).
• dm_from_src() gains dbname argument for MSSQL (#472).

• Fix rows_update() when multiple columns are updated (#488, @samssann).

• enum_fk_candidates() now only checks distinct values, this improves performance for large tables. As a consequence, only the number of distinct values is reported for mismatches, not the number of mismatching rows/entries (#494).

• Fix description of filtering behavior in ?dm_zoom_to (#403).

• Move repository to https://github.com/cynkra/dm (#500).
• Enable more Postgres tests (#497).
• Test DuckDB on GitHub Actions (#498).

• dm_from_src() gains dbname argument for MSSQL (#472).
• Implement count() and tally() for dplyr 1.0.3 compatibility (#475).
• Use databases.pacha.dev instead of db-edu.pacha.dev (#478, @pachamaltese).

• Columns with missing values are no longer primary keys (#469).
• Fix dm_from_src() for MSSQL when learn_keys = FALSE (#427).
• Tests use expect_snapshot() everywhere (#456).
• Fix compatibility with testthat 3.0.1 (#457).

• New vignette("howto-dm-copy", package = "dm") and vignette("howto-dm-rows", package = "dm") discuss updating data on the database. In part derived from vignette("howto-dm-db", package = "dm") (#411, @jawond).
• New dm_mutate_tbl() (#448).
• dm_financial() falls back to db-edu.pacha.dev if relational.fit.cvut.cz is unavailable (#446, @pachamaltese).
• Use testthat 3e (#455).

• Bump RMariaDB required version to 1.0.10 to work around timeout with R CMD check.
• dm_from_src() accepts schema argument for MSSQL databases (#367).

• dm_get_src() returns NULL for local data sources (#394).
• Local target in copy_dm_to() gives a deprecation message (#395).

• copy_dm_to() gives a better error message for bad table_names (#397).
• dm objects with local data sources no longer show the "Table source" part in the output.
• Error messages now refer to "tables", not "elements" (#413).
• New dm_bind() for binding two or more 'dm' objects together (#417).

• For databases, the underlying SQL table names are quoted early to avoid later SQL syntax errors (#419).
• dm_financial() no longer prints message about learn_keys = FALSE.
• dm_rows_update() and related functions now use the primary keys defined in x for establishing matching rows.

• Use withCallingHandlers() where appropriate (#422).
• Consistent definition of .dm and .dm_zoomed methods (#300).
• Examples involving dm_financial() are not run if connection can't be established (#418).
• Fix database tests on CI (#416).

• dm_paste() generates self-contained code (#401).
• Errors regarding cycles in the relationship graph now show the shortest cycle (#405).
• Implement rows_truncate() for databases.
• collect() works on a zoomed dm, with a message.
• The data model is drawn in a more compact way if it comprises of multiple connected components.
• dm_add_pk(check = TRUE) gives a better error message.

• rows_insert() works if column names consist of SQL keywords (#409).
• Cycles in other connected components don't affect filtering in a cycle-free component.
• Avoid src_sqlite() in examples (#372).

• Testing SQLite, Postgres and SQL Server on GitHub Actions (#408, @pat-s).
• Testing packages with all "Suggests" uninstalled.

• New dm_rows_insert(), dm_rows_update(), dm_rows_patch(), dm_rows_upsert(), dm_rows_delete() and dm_rows_truncate(), calling the corresponding rows_*() method for every table (#319).

• New rows_truncate() (#319).

• Added rows_insert() and rows_update() methods for SQLite, Postgres, MariaDB and MSSQL (#319).

• Missing arguments now give a better error message (#388).

• Empty dm object prints as dm() (#386).

• copy_dm_to() also accepts a function as the table_names argument. The unique_table_names() argument is deprecated (#80).

• Add TL;DR to README (#377, @jawond).

• Add content from old README to howto-dm-theory.Rmd (#378, @jawond).

• Require dplyr >= 1.0.0.

• Use GitHub Actions (#369, @pat-s).

• Avoid src_sqlite() in vignettes (#372).
• Rename vignettes (#349).
• Rename error class "dm_error_tables_not_neighbours" to "dm_error_tables_not_neighbors".
• Shortened README and intro article (#192, @jawond).
• Better testing for MSSQL (#339).
• Fix compatibility with dplyr 1.0.0.

• dm_from_src() now works for databases other than Postgres and MSSQL (#288), gives a warning if tables cannot be accessed with table_name = NULL (#348), and gains learn_keys argument to control querying of primary and foreign keys from the database (#340).
• dm_examine_constraints() now prints a different message if a dm has no constraints defined.
• Disambiguation message now only lists column names for easier copy-pasting.
• New methods for "dm_zoomed": head(), tail(), pull(), group_data(), group_indices(), group_vars(), group_keys() and groups() (#236, #203).
• dm_paste() supports writing colors and the table definition via the new options argument. The definition can be written to a file via the new path argument. The select argument is soft-deprecated (#218, #302).
• dm_add_tbl() uses rlang::list2() internally, now accepts := to specify table names.
• New dm_ptype() (#301).
• New dm_financial() and dm_financial_sqlite().
• Printing dm objects from database sources with many tables is now faster (#308, @gadenbuie).
• check_key() now also works on a zoomed dm.
• Key columns are always selected in a join operation, with a message (#153).
• Support alpha colors for the table colors (#279).

• Fix visualization of column that acts as a foreign key more than once (#37).
• dm_add_pk(), dm_rm_pk(), dm_add_fk() and dm_rm_fk() are now stricter when keys exists or when attempting to remove keys that don't exist. A more relaxed mode of operation may be added later (#214).
• examine_cardinality(), dm_examine_constraints() and enum_pk_candidates() now work for columns named n.
• dm_set_key_constraints() (and by extension dm_copy_to(set_key_constraints = TRUE)) now quote identifiers for the SQL that creates foreign keys on the database.
• collect() gives a better error message when called on a "dm_zoomed" (#294).
• check_subset() gives a clean error message if the tables are complex expressions.
• dm_from_src(schema = "...") works on Postgres if search_path is not set on the connection.
• compute.dm_zoomed() no longer throws an error.
• Remove unused DT import (#295).

• Remove use of deprecated src_df() (#336).
• Fix compatibility with dplyr 1.0.0 (#203).

• dm_draw() output is shown in examples (#251).
• New article "{dm} and databases" (#309, @jawond).

• Testing on local data frames (by default), optionally also SQLite, Postgres, RMariaDB, and SQL Server. Currently requires development versions and various pull requests (#334, #327, #312, #76).
• dm_nycflights13(subset = TRUE) memoizes subset and also reduces the size of the weather table.
• Expand definitions of deprecated functions (#204).

• Implement format.dm().
• Adapt to tidyselect 1.0.0 (#257).
• Zooming and unzooming is now faster if no columns are removed.
• Table names must be unique.
• dm_examine_constraints() formats the problems nicely.
• New class for prettier printing of keys (#244).
• Add experimental schema support for dm_from_src() for Postgres through the new schema and table_type arguments (#256).

• Package is now in the "maturing" lifecycle (#154).
• filter.dm_zoomed() no longer sets the filter.
• examine_() functions never throw an error (#238).
• API overhaul: dm_zoom_to(), dm_insert_zoomed(), dm_update_zoomed() and dm_discard_zoomed(); check_() -> examine_(); dm_get_filter() -> dm_get_filters(); dm_from_src() + dm_learn_from_db() -> dm_from_src() (#233).
• New $.dm_zoomed(), [.dm_zoomed(), [[.dm_zoomed(), length.dm_zoomed(), names.dm_zoomed(), tbl_vars.dm_zoomed() (#199, #216).
• New as.list() methods (#213).
• Help pages for dplyr methods (#209).
• New migration guide from dm <= 0.0.5 (#234).
• New {tidyselect} interface for setting colors (#162) and support for hex color codes as well as R standard colors.
• Prepare dm_examine_constraints() and other key-related functions for compound keys (#239).
• Avoid warnings in R CMD check with dev versions of dependencies.
• Improve error messages for missing tables (#220).

• Change cdm_ prefix to dm_. The old names are still available (#117).
• New pull_tbl() extracts a single table from a dm (#206).
• New dm_apply_filters_to_tbl() that applies filters in related tables to a table, similar to dm_apply_filters(); tbl(), $ and [[ no longer apply filter conditions defined in related tables (#161).
• New dm_paste() (#160).
• New check_cardinality() returns the nature of the relationship between parent_table$pk_col and child_table$fk_col (#15).
• New zoom vignette (#171).
• check_key() no longer maps empty selection list to all columns.
• check_key() supports tidyselect (#188).
• dm_rm_tbl() supports tidyselect (#127).
• decompose_table() uses tidyselect (#194).
• Implement copy_to() for dm objects (#129).
• Relax test for cycles in relationship graph (#198).
• Return ref_table column in dm_check_constraints() (#178).
• str() shows simplified views (#123).
• Edits to README (#172, @bbecane).
• Extend validate_dm() (#173).
• Fix zooming into table that uses an FK column as primary key (#193).
• Fix corner case in dm_rm_fk() (#175).
• More efficient check_key() for databases (#208).
• Testing for R >= 3.3 and for debug versions.
• Remove {stringr} dependency (#183).

• cdm_filter() and filter.dm_zoomed() apply the filter instantly, the expression is recorded only for display purposes and for terminating the search for filtered tables in cdm_apply_filters(). This now allows using a variety of operations on filtered dm objects (#124).
• dimnames(), colnames(), dim(), distinct(), arrange(), slice(), separate() and unite() implemented for zoomed dm-s (#130).
• Joins on zoomed dm objects now supported (#121). Joins use the same column name disambiguation algorithm as cdm_flatten_to_tbl() (#147).
• slice.dm_zoomed(): user decides in arg .keep_pk if PK column is tracked or not (#152).
• Supported {dplyr} and {tidyr} verbs are reexported.
• enum_pk_candidates() works with zoomed dm-s (#156).
• New enum_fk_candidates() (#156).
• Add name repair argument for both cdm_insert_zoomed_tbl() and cdm_add_tbl(), defaulting to renaming of old and new tables when adding tables with duplicate names (#132).
• Redesign constructors and validators: dm() is akin to tibble(), dm_from_src() works like dm() did previously, new_dm() only accepts a list of tables and no longer validates, validate_dm() checks internal consistency (#69).
• compute.dm() applies filters and calls compute() on all tables (#135).

• New demo.
• Add explanation for empty dm (#100).

• Avoid asterisk when printing local dm_zoomed (#131).
• cdm_select_tbl() works again when multiple foreign keys are defined between two tables (#122).

• Many {dplyr} verbs now work on tables in a dm. Zooming to a table vie cdm_zoom_to_tbl() creates a zoomed dm on which the {dplyr} verbs can be applied. The resulting table can be put back into the dm with cdm_update_zoomed_tbl() (overwriting the original table) or cdm_insert_zoomed_tbl() (creating a new table), respectively (#89).
• cdm_select_to_tbl() removes foreign key constraints if the corresponding columns are removed.
• Integrate code from {datamodelr} in this package (@bergant, #111).
• Reorder tables in "dm" using cdm_select_tbl() (#108).
• More accurate documentation of filtering operation (#98).
• Support empty dm objects via dm() and new_dm() (#96).
• cdm_flatten_to_tbl() now flattens all immediate neighbors by default (#95).
• New cdm_add_tbl() and cdm_rm_tbl() (#90).
• New cdm_get_con() (#84).
• A dm object is defined using a nested tibble, one row per table (#57).

• cdm_enum_pk_candidates() and cdm_enum_fk_candidates() both show candidates first (#85).
• cdm_flatten_to_tbl() works only in the immediate neighborhood (#75).
• New cdm_squash_to_tbl() implements recursive flattening for left, inner and full join (#75).
• Updated readme and introduction vignette (#72, @cutterkom).
• New cdm_check_constraints() to check referential integrity of a dm (#56).
• cdm_copy_to() gains table_names argument (#79).
• check_key() now deals correctly with named column lists (#83).
• Improve error message when calling cdm_add_pk() with a missing column.

• Fix R CMD check.

• Use caching to improve loading times.
• Run some tests only for one source (#76).
• cdm_enum_fk_candidates() checks for class compatibility implicitly via left_join().
• cdm_enum_fk_candidates() contains a more detailed entry in column why if no error & no candidate (percentage of mismatched vals etc.).
• Improve error messages for cdm_join_to_tbl() and cdm_flatten_to_tbl() in the presence of cycles or disconnected tables (#74).

• Remove the src component from dm (#38).
• Internal: Add function checking if all tables have same src.
• Internal: Add 2 classed errors.
• cdm_get_src() for local dm always returns a src based on .GlobalEnv.
• cdm_flatten() gains ... argument to specify which tables to include. Currently, all tables must form a connected subtree rooted at start. Disambiguation of column names now happens after selecting relevant tables. The resulting SQL query is more efficient for inner and outer joins if filtering is applied. Flattening with a right_join with more than two tables is not well-defined and gives an error (#62).
• Add a vignette for joining functions (#60, @cutterkom).
• Shorten message in cdm_disambiguate_cols().

• cdm_flatten_to_tbl() disambiguates only the necessary columns.
• When flattening, the column name of the LHS (child) table is used (#52).
• Fix formatting in enum_pk_candidates() for character data.
• cdm_add_pk() and cdm_add_fk() no longer check data integrity by default.
• Explicitly checking that the join argument is a function, to avoid surprises when the caller passes data.
• cdm_copy_to() works correctly with filtered dm objects.
• cdm_apply_filters() actually resets the filter conditions.
• A more detailed README file and a vignette for filtering (#29, @cutterkom).
• cdm_draw() no longer supports the table_names argument, use cdm_select_tbl().
• Copying a dm to a database now creates indexes for all primary and foreign keys.

• Requires tidyr >= 1.0.0.
• cdm_nrow() returns named list (#49).
• Remove cdm_semi_join().
• Remove cdm_find_conn_tbls() and the all_connected argument to cdm_select() (#35).
• Unexport cdm_set_key_constraints().
• Rename cdm_select() to cdm_select_tbl(), now uses {tidyselect}.
• cdm_nycflights13() now has cycle = FALSE as default.
• Rename cdm_check_for_*() to cdm_enum_*().

• cdm_filter() only records the filtering operation, the filter is applied only when querying a table via tbl() or when calling compute() or the new cdm_apply_filters() (#32).

• New cdm_flatten_to_tbl() flattens a dm to a wide table with starting from a specified table (#13). Rename cdm_join_tbl() to cdm_join_to_tbl().
• New cdm_disambiguate_cols() (#40).
• New cdm_rename() (#41) and cdm_select() (#50) for renaming and selecting columns of dm tables.
• New length.dm() and length<-.dm() (#53).
• $, [[, [, names(), str() and length() now implemented for dm objects (read-only).
• New enum_pk_candidates().

• browse_docs() opens the pkgdown website (#36).
• as_dm() now also accepts a list of remote tables (#30).
• Use {tidyselect} syntax for cdm_rename_tbl() and cdm_select_tbl() (#14).
• The tibbles returned by cdm_enum_fk_candidates() and cdm_enum_pk_candidates() contain a why column that explains the reasons for rejection in a human-readable form (#12).
• Improve compatibility with RPostgres.
• create_graph_from_dm() no longer fails in the presence of cycles (#10).
• Only suggest {RSQLite}.
• cdm_filter() no longer requires a primary key.
• decompose_table() adds the new column in the table to the end.
• tbl() now fails if the table is not part of the data model.

• Add setup article (#7).

• Using simpler internal data structure to store primary and foreign key relations (#26).
• New nse_function() replaces h() for marking functions as NSE to avoid R CMD check warnings.
• Simplified internal data structure so that creation of new operations that update a dm becomes easier.
• When copying a dm to a database, NOT NULL constraints are set at creation of the table. This removes the necessity to store column types.
• Using {RPostgres} instead of {RPostgreSQL} for testing.

Initial GitHub release.

• dm()
• new_dm()
• validate_dm()
• cdm_get_src()
• cdm_get_tables()
• cdm_get_data_model()
• is_dm()
• as_dm()

• cdm_add_pk()
• cdm_has_pk()
• cdm_get_pk()
• cdm_get_all_pks()
• cdm_rm_pk()
• cdm_check_for_pk_candidates()

• cdm_add_fk()
• cdm_has_fk()
• cdm_get_fk()
• cdm_get_all_fks()
• cdm_rm_fk()
• cdm_check_for_fk_candidates()

• cdm_draw()
• cdm_set_colors()
• cdm_get_colors()
• cdm_get_available_colors()

• cdm_join_tbl()

• cdm_filter()
• cdm_semi_join()
• cdm_nrow()

• cdm_copy_to()
• cdm_set_key_constraints()
• cdm_learn_from_db()

• cdm_is_referenced()
• cdm_get_referencing_tables()
• cdm_select()
• cdm_find_conn_tbls()

• decompose_table()
• reunite_parent_child()
• reunite_parent_child_from_list()

• check_key()
• check_if_subset()
• check_set_equality()
• check_cardinality_0_n()
• check_cardinality_1_n()
• check_cardinality_1_1()
• check_cardinality_0_1()

• cdm_nycflights13()
• cdm_rename_table()
• cdm_rename_tables()