A hex logo has been added to the package (@edgararuiz, #824).
databricks() and snowflake() detect
viewer-based credentials on Posit Connect (@atheriel, #853 and #894).
odbcEditDrivers(), odbcEditSystemDSN(),
and odbcEditUserDSN() create shortcuts for
file.edit(odbcListConfig()[[i]]) on macOS and Linux (@simonpcouch,
#827).
New redshift() helper simplifies connections to
Amazon Redshift clusters, particularly with IAM credentials (@atheriel, #879).
Error messages rethrown from drivers no longer become garbled when raw messages contain curly brackets (@simonpcouch, #859).
SQL Server drivers installed with homebrew are now discovered automatically on ARM macOS (@stevecondylios, #893).
databricks(workspace) removes trailing slashes
automatically, preventing uninformative driver messages (@simonpcouch,
#827).
dbConnect() displays clear errors when unixODBC
isn’t found on macOS and Linux (@simonpcouch, #782).
dbConnect() adds name_encoding to
complement encoding for separate column content and name
encoding (#845).
dbListTables() and dbExistsTable()
improve temp table support with DB2 (#823).
Netezza character and boolean columns receive improved data type inference (#847, #850).
Oracle DATE and TIMESTAMP(n) targets
handle writes correctly with batch_size > 1
(#810).
snowflake() checks runtime driver configuration on
macOS (#857).
snowflake() accepts uid without
pwd when using
authenticator = "externalbrowser" or
authenticator = "SNOWFLAKE_JWT" (@simonpcouch, #817 and #889).
Resolved SQL Server data truncation in temp tables with
FreeTDS (#866), writing issues with SIMBA drivers (#816),
hms data roundtrip (#887), and precision when writing to
DATETIME2 targets (#793).
New function snowflake() makes it easier to connect
to Snowflake, automatically handling authentication correctly on
platforms that provide Snowflake-native OAuth credentials (@atheriel, #662).
Long running queries can now be interrupted using Ctrl-C. This
feature is enabled by default in interactive sessions. It can be
controlled by the interruptible argument to
dbConnect() or by the global option
odbc.interruptible. Should be considered experimental—if
you experience problems please file an issue on the package’s GitHub
repository (#796).
Improved argument checking and transitioned to the cli package for formatting most existing error messages (@simonpcouch, #781, #784, #785, #788).
Raises “Cancelling previous query” warnings from R rather than
from Rcpp when a connection has a current result to avoid possible
incorrect resource unwinds with options(warn = 2)
(#797).
Adjusted the default batch_rows value for
dbWriteTable() and dbBind() methods. odbc
1.3.0 changed the default value from 1024 to NA, which sets
the batch size to be the length of the input. While this addressed
issues for some drivers when 1024 was greater than the number of rows,
it also led to excessive memory consumption when the input was very
large for some other drivers. The package will now interpet
NA as the minimum of 1024 and the length of the input
(@simonpcouch,
#774).
The encoding of non-ASCII column names of SQL results is now always converted to UTF-8. (@shrektan, #430)
Improved error messages when the encoding of client and db-server are different. (@shrektan, #432)
dbListFields() now works with Id() and
SQL() identifiers (#771).
Transitioned odbcDataType() to use S4 for
consistency. S3 methods defined locally will need to be rewritten (@simonpcouch,
#701).
The "OdbcConnection" method for
dbQuoteIdentifier() will no longer pass x to
encodeString() before returning, for consistency with the
default implementation in DBI (@simonpcouch, #765).
A bug in the implementation of a new feature introduced in 1.4.2,
where the package would automatically set the ODBCSYSINI
environmental variable when using the unixODBC driver manager, was
fixed; that environmental variable will now actually be set on package
load (@simonpcouch, #792).
databricks() will now automatically configure the
needed driver and driver manager on macOS (@simonpcouch, #651).
databricks() now picks up on Posit Workbench-managed
Databricks credentials when rendering Quarto and RMarkdown documents in
RStudio (@atheriel,
#805).
Improved performance on write with Snowflake (#760).
Resolved issue when previewing tables using the RStudio Connections pane with Teradata (@simonpcouch, #755).
dbAppendTable() Improve performance by checking
existence once (#691).
dbConnect() no longer automatically escapes
suspicious characters (since there doesn’t seem to be a consistent way
to do this across drivers) but instead points you to
quote_value() which applies a heuristic that should work
for most drivers (#718).
New wrapper for dbExecute() that sets
immediate = TRUE if you are not supplying
params. That should yield a small speed boost in many cases
(#706).
dbSendQuery() once again defaults to
immediate = FALSE (since if you’re using it instead of
dbGetQuery() you’re likely to be using it with
dbBind()). (#726).
Deprecated odbcConnectionColumns() (in favor of
dbListFields()), odbcConnectionActions(), and
odbcConnectionIcon() (@simonpcouch, #699).
Backend specific changes:
databricks: Fix schema enumeration in connections pane (@detule, #715).
Oracle: use more reliable technique to determine user/schema name
(#738), and fix dbExistsTable() when identifier components
contain _ (@detule, #712).
SQL Server: improvements to dbExists() (@meztez, #724) and
dbListTables() (@simonpcouch, #509) for temporary
tables. It now uses column type "BIGINT" integer64
objects.
SQL Server with freetds driver: no longer crashes when executing multiple queries (@detule, #731).
Teradata: Fix usage of exact argument in internal
methods (@detule,
717).
On MacOS and Linux, the package will now automatically set the
ODBCSYSINI environmental variable when using the unixODBC
driver manager. ODBCSYSINI will not be changed if it exists
already (@simonpcouch, #709).
New odbcListConfig() lists configuration files on
Mac and Linux (@simonpcouch, #565).
databricks() now works with manually supplied
pwd and uid (#690).
Oracle: uses correct parent class (#685).
SQL Server: correctly enumerate schemas across databases in connections pane (@detule, #527).
New odbc::databricks() makes it easier to connect to
Databricks, automatically handling many common authentication scenarios
(@atheriel,
#615).
dbListTables(), dbListFields() and
dbExistsTable() automatically escape underscores in
identifier arguments. This leads to substantial performance improvements
for some backends (e.g. snowflake) (@detule, @fh-afrachioni, #618).
dbGetQuery() and dbSendQuery() now set
immediate = TRUE if you are not using a parameterised
query. That should yield a small speed boost in many cases
(#633).
Increased the minimum required R version from 3.2.0 to 3.6.0 (@simonpcouch, #629).
S4 classes for the most database drivers are now exported, make it possible to use in other packages (#558).
ODBC errors are now spread across multiple lines, making them easier to read (@detule, #564).
DBI::dbConnect(odbc::odbc()) now gives a clear error
if you supply multiple arguments with the same name when case is ignored
(#641).
DBI::dbConnect(odbc::odbc()) now automatically
quotes argument values that need it (#616).
Oracle: Fix regression when falling back to
odbcConnectionColumns() to describe column data types
(@detule,
#587)
Spark SQL: Correctly enumerate schemas away from the current catalog (@detule, #614)
Snowflake: improved translation from R to snowflake types (@meztez, #599).
SQL Server
Improved handling for local temp tables in
dbWrite(), dbAppendTable(), and
dbExistTable() (@detule, #600)
Specialize syntax in sqlCreateTable to avoid
failures when writing to (new) local temp tables. (@detule, #601)
Teradata: Improved handling for temp tables (@detule and @But2ene, #589, 590)
Various fixes for R CMD check.
The argument order for sqlCreateTable() methods
avoiding weird argument mismatch problems (#448).
Fixed truncation when retrieving unicode data stored in VAR/CHAR columns (@detule, #553).
Fixed issue related to fetching zero rows (@detule, #528).
Backend specific improvements:
DATE to
TIMESTAMP (@schuemie, #555).Optimized table preview methods. Enhances large table preview within RStudio IDE (@detule, #525).
Added attributes parameter to
dbConnect(...). Can be used to pass a token and
authenticate against Azure AD when using Microsoft SQL Server.
Documented in ?ConnectionAttributes (@detule, #521).
Fix length(x) = 3 > 1' in coercion to 'logical(1)
warning in connection observer (@meztez, #494).
odbcDataType.Snowflake() method for Snowflake
databases. (@edgararuiz, #451)invalid descriptor issues when retrieving results
from SQL Server + Microsoft’s ODBC driver, using parametrized queries.
(@detule, #414)dbQuoteString now respects this when called on a connection
of class Hive. (@rnorberg, #184)sqlCreateTable(con, ..., temporary = TRUE)
and con is a connection of class DB2/AIX64,
the CREATE TABLE statement that is generated properly
creates a temporary table in DB2. The statement begins with
DECLARE GLOBAL TEMPORARY TABLE at
https://www.ibm.com/docs/SSEPEK_11.0.0/sqlref/src/tpc/db2z_sql_declareglobaltemptable.html
and ends with ON COMMIT PRESERVE ROWS (DB2’s default
behavior is ON COMMIT DELETE ROWS, which results in the
inserted data being deleted as soon as dbWriteTable
completes). (@rnorberg, #426)invalid descriptor errors from drivers such as
Microsoft SQLServer driver and the freeTDS driver which do not support
out of order retrieval are now avoided. This is done by unbinding any
nanodbc buffer past the long column. Performance for the unbound columns
in these cases will be reduced, but the retrieval will work without
error (@detule,
#381)dbBind() and dbFetch() now support
multiple result sets (@vkapartzianis, #345)dbAppendTable() method for OdbcConnection objects
(#335)dbQuoteIdentifier() now uses the input names (if
any).dbWriteTable() and dbBind() now default to
a batch_rows of NA, which sets the batch size
to be the length of the input. This avoids problems with drivers that
don’t support batch sizes larger than the input size. To restore the
behavior prior to this release pass batch_rows = 1024 or
set options(odbc.batch_rows = 1024) (#391).dbWriteTable() now handles
data.table::IDate() objects (#388)dbWriteTable(field.types=) now issues a warning rather
than an error for missing columns (#342)odbcConnectionColumns() now works as intended with
DBI::Id() objects (#389)dbFetch() now verifies that n is a valid
input.timezone parameter (#398)Date and
POSIXct types via DATE and
TIMESTAMP data types (#324, #349, #350)dbWriteTable() now executes immediately, which fixes
issues with temporary tables and the FreeTDS SQL Server driver (@krlmlr).
blob roundtrips now work in more cases (#364, @krlmlr).
The BH dependency has been removed (#326)
The configure script now again uses iodbc-config on macOS, if available.
Fix errors with drivers who do not fully implement SQLGetInfo (#325).
Fix the unicode character width issues with the macOS odbc CRAN binary. (#283)
The configure script now works on systems whose /bin/sh does not support command substitution, such as Solaris 10.
sqlCreateTable() and dbWriteTable() now
throw an error if you mistakenly include field.types names
which do not exist in the input data. (#271)
The error message when trying to write tables with unsupported types now includes the column name (#238).
dbConnect() now has a new param
timezone_out which is useful if the user wants the datetime
values be marked with a specific timezone instead of UTC (@shrektan, #294).
dbGetQuery(), dbSendQuery() and
dbSendStatement() gain a immediate argument to
execute the statement or query immediately instead of preparing, then
executing the statement. (#272, @krlmlr)
dbGetQuery(), dbSendQuery() and
dbSendStatement() gain a params argument,
which allows them to be used (indirectly) by
DBI::dbAppendTable() (#210, #215, #261).
dbWriteTable() and dbBind() methods
gain a batch_rows argument, to control how many rows are
bound in each batch. The default can be set globally with
options(odbc.batch_rows). This is useful if your database
performs better with a larger or smaller number of rows per batch than
the default of 1024. (#297)
New odbcConnectionColumns() function to describe the
column types. This data is used when binding columns, which allows
drivers which do not support the SQLDescribeParam() ODBC
function, such as freeTDS to work better with bound columns. (#313,
@detule)
Added a Teradata odbcDataType() to support writing
logical data to Teradata servers (#240, @blarj09).
Added a Access odbcDataType() method to support
writing to Access databases (#262, @vh-d)
odbcListDrivers() gains a keep and
filter argument and global options
odbc.drivers_keep, odbc.drivers_filter to keep
and filter the drivers returned. This is useful if system administrators
want to reduce the number of drivers shown to users. (@blairj09, #274)
Subseconds are now retained when inserting POSIXct objects (#130, #208)
The RStudio Connections Pane now shows the DSN, when available (#304, @davidchall).
SQL Server ODBC’s now supports the ‘-155’ data type, and its
losing sub-second precision on timestamps; this still returns type
DATETIMEOFFSET as a character, but it preserves sub-seconds
and has a numeric timezone offset (@r2evans, #207).
dbExistsTable() now handles the case-sensitivity
consistently as other methods (@shrektan, #285).
dbExistsTable() now works for SQL Server when
specifying schemas but not catalogs using the freeTDS and Simba drivers.
(#197)
DBI::dbListFields() no longer fails when used with a
a qualified Id object (using both schema and table) (#226).
dbWriteTable() now always writes
NA_character as NULL for data.frame with only
one row (@shrektan,
#288).
Fix an issue that the date value fetched from the database may be one day before its real value (@shrektan, #295).
dbConnect() gains a timeout parameter,
to control how long a connection should be attempted before timing out
(#139).
Full schema support using Id objects available in DBI 0.8 (#91, #120).
dbQuoteIdentifier() when given
a length 0 input (#146 @edgararuiz).Add custom sqlCreateTable and
dbListTables method for Teradata connections (#121, @edgararuiz).
Add dbms.name parameter to dbConnect()
to allow the user to set the database management system name in cases it
cannot be queried from the Driver (#115, @hoxo-m).
Fix multiple transactions with rollback (#136).
Fix translation of missing values for integer types (#119).
Update PIC flags to avoid portability concerns.
Add bigint parameter to dbConnect() to
allow the user to set the behavior when converting 64 bit integers into
R types.
Fixes for compatibility with the RStudio Connections pane and viewer. (@jmcphers).
Define BUILD_REAL_64_BIT_MODE for compatibility with
older systems sql.h, to ensure we always use 8 byte
pointers.
Added temporary-table support for Oracle database using a custom
sqlCreateTable (#99, @edgararuiz).
Fix regression when binding due to the num_columns variable not
being updated by odbc_result::bind_list().
Support table creation for Vertica and Redshift (#93, @khotilov).
Changed parameter fieldTypes to
field.types in functions dbWriteTable() and
sqlCreateTable() to be compliant with DBI (#106, @jschelbert).
dbSendStatement no longer executes the statement, this is instead
done when dbBind() or dbGetRowsAffected() is
called. This change brings ODBC into compliance with the DBI
specification for dbSendStatement() (#84, @ruiyiz).
Workaround for drivers which do not implement SQLGetInfo, such as the Access driver (#78).
Fix for installation error for systems without GNU Make as the default make, such as Solaris.
Provide a fall backs for drivers which do not support
SQLDescribeParam and those which do not support
transactions.
sqlCreateTable() gains a fieldTypes
argument, which allows one to override a column type for a given table,
if the default type is not appropriate.
Support for databases with non UTF-8 encodings. Use the
encoding parameter to dbConnect() to specify
the database encoding.
Support for the SQL Connection Pane in the RStudio IDE.
Unknown fields no longer throw a warning, instead they signal a custom condition, which can be caught if desired with
withCallingHandlers(expr, odbc_unknown_field_type = function(e) {…} )
Conversion to and from timezones is handled by the cctz library. If the database
is in a non-UTC timezone it can be specified with the
timezone parameter to dbConnect().
Time objects are converted to and from hms
objects.
64 bit integers are converted to and from bit64
objects.
Support table creation for Impala and Hive Databases (#38, @edgararuiz).
NEWS.md file to track changes to the
package.