Breaking changes

Bug fixes

  • Use NULL in favor of NULL::text when quoting strings and literals, to support JSON and other text-ish types. Fixes a regression introduced in #370 (#393, #425).

Features

Libraries

  • Windows: update to libpq-15.3 (#442).

  • Upgrade boost to 1.81.0-1 to fix sprintf warnings (#417).

Documentation

  • Suppress warning in gcc-12 (#443).

  • Tweak driver docs (@dpprdan, #433).

  • Relicense as MIT.

Testing

  • Close result set.

Internal

Features

Bug fixes

Internal

  • Expand tests for dbConnect(check_interrupts = TRUE) (#385, @zozlak).
  • Ignore extended timestamp tests on i386 (#387).

Features

Internal

  • Make method definition more similar to S3. All setMethod() calls refer to top-level functions (#380).

Features

  • dbWriteTable() uses savepoints for its transactions, even if an external transaction is open. This does not affect Redshift, because savepoints are not supproted there (#342).
  • With dbConnect(check_interrupts = TRUE), interrupting a query now gives a dedicated error message. Very short-running queries no longer take one second to complete (#344).

Bug fixes

  • dbQuoteLiteral() correctly quotes length-0 values (#355) and generates typed NULL expressions for NA values (#357).
  • The SET DATESTYLE query sent after connecting uses quotes for compatibility with CockroachDB (#360).

Internal

  • dbConnect() executes initial queries with immediate = TRUE (#346).
  • Check Postgres starting from version 10 on GitHub Actions (#368).
  • Fix build on Ubuntu 16.04 (#352).
  • Mention libssl-dev in configure script (#350).

Bug fixes

  • Avoid crash by dereferencing 0-size vector (#343).

Features

  • Redshift() connections now adhere to almost all of the DBI specification when connecting to a Redshift cluster. BLOBs are not supported on Redshift, and there are limitations with enumerating temporary and persistent tables (#215, #326).
  • dbBegin(), dbCommit() and dbRollback() gain name argument to support savepoints. An unnamed transaction must be started beforehand (#13).
  • dbWriteTable() uses a transaction (#307).
  • dbSendQuery() gains immediate argument. Multiple queries (separated by semicolons) can be passed in this mode, query parameters are not supported (#272).
  • dbConnect(check_interrupts = TRUE) now aborts a running query faster and more reliably when the user signals an interrupt, e.g. by pressing Ctrl+C (#336).
  • dbAppendTable() gains copy argument. If set to TRUE, data is imported via COPY name FROM STDIN (#241, @hugheylab).
  • Postgres NOTICE messages are now forwarded as proper R messages and can be captured and suppressed (#208).

Bug fixes

  • dbQuoteLiteral() converts timestamp values to input time zone, used when writing tables to Redshift (#325).

Internal

Documentation

  • Order help topics on pkgdown site.
  • Use @examplesIf in method documentation.
  • Document when field.types is used in dbWriteTable() (#206).
  • Document setting the tablespace before writing a table (#246).
  • Tweak error message for named params argument to dbBind() (#266).
  • Fix dbConnect(check_interrupts = TRUE) on Windows (#244, @zozlak).
  • Windows: update to libpq 13.2.0 and add UCRT support (#309, @jeroen).
  • Remove BH dependency by inlining the header files (#300).
  • Use Autobrew if libpq is older than version 12 (#294, @jeroen).
  • dbConnect() now issues SET datestyle to iso, mdy to avoid translation errors for datetime values with databases configured differently (#287, @baderstine).

Bug fixes

  • Inf, -Inf and NaN values are returned correctly on Windows (#267).
  • Fix behavior with invalid time zone (#284, @ateucher).

Internal

  • dbConnect() defaults to timezone_out = NULL, this means to use timezone.
  • FORCE_AUTOBREW environment variable enforces use of autobrew in configure (#283, @jeroen).
  • Fix configure on macOS, small tweaks (#282, #283, @jeroen).
  • Fix configure script, remove $() not reliably detected by checkbashisms.
  • configure uses a shell script and no longer forwards to src/configure.bash (#265).

Bug fixes

  • Inf, -Inf and NaN values are returned correctly on Windows (#267).
  • DATETIME values (=without time zone) and DATETIMETZ values (=with time zone) are returned correctly (#190, #205, #229), also if they start before 1970 (#221).

Internal

  • configure uses a shell script and no longer forwards to src/configure.bash (#265).
  • Switch to GitHub Actions (#268, thanks @ankane).
  • Now imports the lubridate package.
  • Gains new Redshift driver for connecting to Redshift databases. Redshift databases behave almost identically to Postgres so this driver allows downstream packages to distinguish between the two (#258).
  • Datetime values are now passed to the database using an unambiguous time zone format (#255, @imlijunda).
  • Document Postgres() together with dbConnect() (#242).
  • Windows: update libpq to 12.2.0.

Communication with the database

  • Breaking: Translate floating-point values to DOUBLE PRECISION by default (#194).
  • Avoid aggressive rounding when passing numeric values to the database (#184).
  • Avoid adding extra spaces for numerics (#216).
  • Column names and error messages are UTF-8 encoded (#172).
  • dbWriteTable(copy = FALSE), sqlData() and dbAppendTable() now work for character columns (#209), which are always converted to UTF-8.

New features

  • Add timezone argument to dbConnect() (#187, @trafficonese).
  • Implement dbGetInfo() for the driver and the connection object.
  • dbConnect() gains check_interrupts argument that allows interrupting execution safely while waiting for query results to be ready (#193, @zozlak).
  • dbUnquoteIdentifier() also handles unquoted identifiers of the form table or schema.table, for compatibility with dbplyr. In addition, a catalog component is supported for quoting and unquoting with Id().
  • dbQuoteLiteral() available for "character" (#209).
  • Windows: update libpq to 11.1.0.
  • Fulfill CII badge requirements (#227, @TSchiefer).

Bug fixes

  • Hide unused symbols in shared library (#230, @troels).
  • Fix partial argument matching in dbAppendTable() (r-dbi/DBI#249).
  • Fix binding for whole numbers and POSIXt timestamps (#191).

Internal

  • Replace std::mem_fn() by boost::mem_fn() which works for older compilers.
  • Replace std::mem_fun_ref() by std::mem_fn().
  • Add support for bigint argument to dbConnect(), supported values are "integer64", "integer", "numeric" and "character". Large integers are returned as values of that type (r-dbi/DBItest#133).
  • Data frames resulting from a query always have unique non-empty column names (r-dbi/DBItest#137).
  • New arguments temporary and fail_if_missing (default: TRUE) to dbRemoveTable() (r-dbi/DBI#141, r-dbi/DBI#197).
  • Using dbCreateTable() and dbAppendTable() internally (r-dbi/DBI#74).
  • The field.types argument to dbWriteTable() now must be named.
  • Using current_schemas(true) also in dbListObjects() and dbListTables(), for consistency with dbListFields(). Objects from the pg_catalog schema are still excluded.
  • dbListFields() doesn’t list fields from tables found in the pg_catalog schema.
  • The dbListFields() method now works correctly if the name argument is a quoted identifier or of class Id, and throws an error if the table is not found (r-dbi/DBI#75).
  • Implement format() method for SqliteConnection (r-dbi/DBI#163).
  • Reexporting Id(), DBI::dbIsReadOnly() and DBI::dbCanConnect().
  • Now imports DBI 1.0.0.
  • Breaking change: dbGetException() is no longer reexported from DBI.
  • Make “typname” information available after dbFetch() and dbGetQuery(). Values of unknown type are returned as character vector of class "pq_xxx", where "xxx" is the “typname” returned from PostgreSQL. In particular, JSON and JSONB values now have class "pq_json" and "pq_jsonb", respectively. The return value of dbColumnInfo() gains new columns ".oid" (integer), ". known" (logical) and ".typname" (character) (#114, @etiennebr).
  • Values of class "integer64" are now supported for dbWriteTable() and dbBind() (#178).
  • Schema support, as specified by DBI: dbListObjects(), dbUnquoteIdentifier() and Id().
  • Names in the x argument to dbQuoteIdentifier() are preserved in the output (r-lib/DBI#173).
  • All generics defined in DBI (e.g., dbGetQuery()) are now exported, even if the package doesn’t provide a custom implementation (#168).
  • Replace non-portable timegm() with private implementation.
  • Correct reference to RPostgreSQL package (#165, @ClaytonJY).
  • Only call PQcancel() if the query hasn’t completed, fixes transactions on Amazon Redshift (#159, @mmuurr).
  • Fix installation on macOS.
  • Check libpq version in configure script (need at least 9.0).
  • Fix UBSAN warning: reference binding to null pointer (#156).
  • Fix rchk warning: PROTECT internal temporary SEXP objects (#157).
  • Fix severe memory leak when fetching results (#154).

Initial release, compliant to the DBI specification.

  • Test almost all test cases of the DBI specification.
  • Fully support parametrized queries.
  • Spec-compliant transactions.
  • 64-bit integers are now supported through the bit64 package. This also means that numeric literals (as in SELECT 1) are returned as 64-bit integers. The bigint argument to dbConnect() allows overriding the data type on a per-connection basis.
  • Correct handling of DATETIME and TIME columns.
  • New default row.names = FALSE.