CHANGELOG.md

Changelog

10.0.0 - 2024-11-24

Added

• [breaking] odbc-api will now use narrow function calls by default on non-windows systems. This assumes that the ODBC driver on that platform uses UTF-8 encoding. This is usually the case as many Linux systems use an UTF-8 locale. Outside of windows the wide function calls are usually less battle tested on drivers. Downstream artefacts like arrow-odbc and odbc2parquet therefore have been compiling with the narrow flag on non-windows systems for quite a while now to opt into the behavior which most likely works by default. However if somebody had a non-windows platform with a non-utf 8 local and a driver which actually could use UTF-16, he could not set appropriate compiler flags to revert the narrow feature added by these crates. Since the default behavior for each platform is now triggered by odbc-api itself, downstream artefacts can overwrite it both ways. E.g. using the narrow flag on windows to get some speed, if they know their target platform has a UTF-8 local configured. Or the other way around using the wide flag on a Linux system to e.g. handle special character in column names, which only seem to work with the wide variants of the drivers. So if you used odbc-api previously on Linux without any flags it did by default use the wide function calls. It now uses the narrow ones by default. If you want the old behavior just specify the wide feature flag. For windows users nothing breaks.

Other

• (deps) bump thiserror from 2.0.0 to 2.0.3
• (deps) bump csv from 1.3.0 to 1.3.1
• (deps) bump tempfile from 3.13.0 to 3.14.0
• (deps) bump tokio from 1.41.0 to 1.41.1
• (deps) bump thiserror from 1.0.68 to 2.0.0
• (deps) bump anyhow from 1.0.92 to 1.0.93
• (deps) bump thiserror from 1.0.66 to 1.0.68
• (deps) bump thiserror from 1.0.65 to 1.0.66
• (deps) bump anyhow from 1.0.91 to 1.0.92
• (deps) bump syn from 2.0.87 to 2.0.89
• (deps) bump syn from 2.0.86 to 2.0.87
• (deps) bump syn from 2.0.85 to 2.0.86

9.0.0

• ConcurrentBlockCursor is now generic over the buffer type

8.1.4

• Fix failed release 8.1.3 due to out of sync version for derive package

8.1.3

• Fix: Failed attempts to rollback transactions during Drop of a connection will no longer panic.

8.1.2

• Fix: Connection::execute_polling did not enable asynchronus execution on the statement handle. This means that the operations on that cursor were actually blocking even if the driver does support polling.

8.1.1

• Fix release. Release from main branch

8.1.0

• ConnectionAndError now implements std::error::Error.

8.0.1-2

• Fix: Fix version of odbc-api-derive crate

8.0.0

• Connection::into_cursor now allows you to recover the Connection in case of failure. For this the return type has been changed from Error to ConnectionAndError. Error implements From<ConnectionAndError>, so converting it to Error via ? is frictionless.

7.2.4

• Improved documentation

7.2.3

• Update dependencies

7.2.2

• Fix: Add metadata for odbc-api-derive crate

7.2.1

• Fix: Release pipeline for derive feature

7.2.0

• Adds derive feature for deriving custom implementations of FetchRow.

7.1.0

• Adds support for row wise bulk fetching via RowVec.

7.0.0

• Removed Quirks. Detection of driver problems via dbms name is not relyable enough.
• Support setting packet size, via ConnectionOptions::packet_size.

6.0.2

• Fixes an issue with decimal_text_to_i128, which caused a negative value to be treated as positive, if it had been below a magnitude of one. Thanks to @timkpaine for reporting and fixing the issue.

6.0.1

• Fixes an issue with Environment::drivers, which caused it to have at most one driver attribute in the DriverInfo attributes.

6.0.0

• ConcurrentBlockCursor::from_block_cursor is now infalliable.
• Quirks::indicators_returned_from_bulk_fetch_are_memory_garbage will be now set for every database name which starts with DB2.

5.1.0

• Adds ConcurrentBlockCursor to ease fetching large result sets in a dedicated system thread.

5.0.0

• removed reexport of force_send_sync. It is no longer used within odbc-api this should have been removed together with the methods which were promoting Connections to Send, but has been overlooked.
• Adds decimal_text_to_i128 a useful function for downstream applications including odbc2parquet and arrow-odbc.

4.1.0

• Introducing Quirks to document knoweldege about behaviour from drivers which is different from the ODBC specification. This can be helpful for generic applications in order to support databases if they want to invest in workarounds for bugs or percularities of ODBC drivers.

4.0.1

• Additional debug log messages around allocating and deallocating handles.

4.0.0

• ResultSetMetadata::col_display_size now returns Option<NonZeroUsize> instead of isize.
• ResultSetMetadata::col_octet_length now returns Option<NonZeroUsize>, instead of isize.
• ResultSetMetadata::utf8_display_sizes now iterates over Option<NonZeroUsize>, instead of usize.
• DataType now stores the length of variadic types as Option<NonZeroUsize> insteaf of usize.
• DataType::display_size now returns Option<NonZeroUSize> instead of Option<usize>.
• DataType::utf8_len now returns Option<NonZeroUSize> instead of Option<usize>.
• DataType::utf16_len now returns Option<NonZeroUSize> instead of Option<usize>
• DataType::column_size now returns Option<NonZeroUSize> instead of usize
• BufferDesc::from_data_type now returns None for variadic types without upper bound instead of a zero sized buffer.
• Indicator::value_len now is named Indicator::length.
• Provide buffer index for truncation errors

3.0.1

• Setting only odbc_version_3_5 without deactivating default features now compiles again. It is still recommended though to use the odbc_version_3_5 flag only in combination with --no-default-features since odbc_version_3_8 is a default feature.

3.0.0

• ParameterDescription::nullable has been renamed to ParameterDescription::nullability
• Remove deprecated function Connection::promote_to_send.
• ErrorUnsupportedOdbcVersion is now returned if driver manager emits a diagnostic record using with status S1009 then setting ODBC version. That status code used to mean "Invalid Attribute" back in the days of ODBC 2.x.

2.2.0

• Adds BlockCursor::row_array_size, in order to infer maximum batch size without unbinding the row set buffer first. This allows downstream arrow-odbc for a faster implementation then creating a double buffered concurrent stream of batches directly from a block cursor.

2.1.0

It is now better possible to use odbc-api with multithreading in completly safe code.

• Connection is now Send.
• StatementConnection is now Send.

According to the standard this is guaranteed to be safe. By now this has also worked in practice long enough and with enough different drivers, that this can be considered safe, not only by the ODBC standard, but also in practice.

2.0.0

• Variant Error::TooLargeValueForBuffer now has a member indicator, indicating the size of the complete value in the DBMS. This has been introduced to help users decide how to pick a maximum string size in bulk fetches, in the presence of large variadic columns.

1.0.2

• Fix: Cursor::fetch_with_truncation_check did erroneously report truncation errors for NULL fields, if the query also triggered any ODBC diagnostic. This has been fixed in the previous release for variadic binary column buffers but a similar failure had been missed for variadic text columns.

1.0.1

• Fix: Cursor::fetch_with_truncation_check did erroneously report truncation errors for NULL fields, if the query also triggered any ODBC diagnostic. This remained undedected for a while, because in every situation that there is a truncation, there must also be a diagnostic according to the ODBC standard. However the reverse is not true. In this situation a NULL had been erronously treated the same as a NO_TOTAL indicator and a false positive has been reported to a user claiming that there is a truncation, there in fact is none.

1.0.0

• Introduced CursorRow::get_wide_text in order to fetch large UTF-16 values into a buffer.
• The IntoParameter implementation for str slices and Strings will now be converting them to VarWCharBox if the narrow compile time feature is not activated. This has been done in order to allow for easier writing of portable code between Windows and non-Windows platforms. Usually you would like to activate the narrow feature on non-windows platform there a UTF-8 default locale can be assumed. On Windows you want utilize UTF-16 encoding since it is the only reliable way to to transfer non ASCII characters independent from the systems locale. With this change on both platforms one can now simply write:

// Uses UTF-16 if `narrow` is not set. Otherwise directly binds the UTF-8 silce.
conn.execute(&insert_sql, &"Frühstück".into_parameter()).unwrap();

// Guaranteed to use UTF-16 independent of compiliation flags.
let arg = U16String::from_str("Frühstück");
conn.execute(&insert_sql, &arg.into_parameter()).unwrap();

// Guaranteed to use UTF-8 independent of compiliation flags. This relies on the ODBC driver to use
// UTF-8 encoding.
let arg = Narrow("Frühstück");
conn.execute(&insert_sql, &arg.into_parameter()).unwrap();

0.57.1

• Corrected typos in documentation. Thanks to @zachbateman

0.57.0

• Support for U16String and Option<U16Str> as input parameter via IntoParameter trait.
• Added VarWCharSlice, VarWCharBox, VarWCharSliceMut, VarWCharArray.

0.56.2

• Support U16Str and Option<U16Str> as input parameter via IntoParameter trait.

0.56.1

• Fix: Statement::complete_async is now annotated to be only available in ODBC version 3.8. This missing annotation prevented compilation then specifying ODBC version 3.5.

0.56.0

• CursorRow::get_variadic is now private as it was always intended to be a private implemenatation detail.
• Statement::set_num_rows_fetched has been split into Statement::set_num_rows_fetched and Statement::unset_num_rows_fetched in order to reduce the amount of unsafe code a bit.
• Adds Cursor::more_results to support statements which return multiple result sets.

0.55.0

• Added support for login timeout, i.e. the user can now explicitly specify a time in seconds after which a login attempt to a datasource should fail. This triggered two breaking changes:

• Environment::connect_with_connection_string now takes an additional argument ConnectionOptions.

• Environment::connect now takes an additional argument ConnectionOptions.

Migration is trivial, by supplying ConnectionsOptions::default() as the last argument.

0.54.3

• Added catalog function foreign_keys to Connection and Preallocated.

0.54.2

• CursorRow::get_text or Cursor::get_binary now panics if the SQLGetData implementation of the ODBC driver reports inconsistent indicator for the same cell. Previously it could have looped forever.

0.54.1

• Introduced common implementation VarCell for VarBinary and VarChar.

0.54.0

• Changed checking of truncated values after fetch. The new method checks the indicator buffer, rather than relying on diagnostics. This means Error::TooManyDiagnostics can no longer occurr.

0.53.0

• Better error messages if checking for truncation fails due to too many diagnostics
• Removed deprecated
  • BufferDescription
  • BufferKind
• Removed deprecated names
  • AnyColumnBuffer
  • AnyColumnSliceMut
  • AnyColumnView
  • RowSetCursor
  • RowSetCursorPolling

0.52.4

• Improved support for fetching parameter descriptions of prepared statements:
  • Introduced method Prepared::num_params
  • Introduced method Prepared::parameter_descriptions

0.52.3

• Introduced method Prepared::column_inserter to replace Prepared::any_column_inserter.
• Introduced method Prepared::into_column_inserter to replace Prepared::into_any_column_inserter.

0.52.2

• Introduced associated method Item::buffer_desc to replace Item::BUFFER_KIND.

0.52.1

• BufferKind and BufferDescription are now unified to BufferDesc.
• Renamed RowSetCursor into BlockCursor.
• Fix: ColumnarBulkInserter::new now resets all parameters on the statement handle in case it fails. This prevents accessing the bound parameters on a borrowed statement handle after the constructor failed, at which point their valididty is no longer guaranteed.

0.52.0

• sys::Timestamp is now directly supported with get_data.
• InputParameter is now implying the new CElement trait.

0.51.0

• Removed ColumnProjection trait.
• ColumnBuffer now has generic associated type View instead of inheriting from ColumnProjection.
• WithDataType can now also be used to annotate the Relational type on Timestamp.
• WithDataType<Timestamp> now implements OutputParameter.
• Minimum Rust version is 1.65.0

0.50.3

• Fix: There has been an issue with fetching text values using get_text from a db2 System. db2 is not part of the test suite (yet?), so not entirely sure that it is truly gone yet, but chances are its fixed.
• Implement ColumnarBuffer for Vec<T>.

0.50.2

• TextBufferRowSet::for_cursor now applies the upper bound column size limits also to buffers if the metadata indicates a size of 0. This allows e.g. fetching values from Microsoft SQL Server columns with type VARCHAR(max).

0.50.1

• Updated documentation for BufferDescription and DataType.

0.50.0

• Fetching diagnostic records now starts with a message buffer size of 512 instead of 0 in order to recduce calls to SQLGetDiagRec.

Breaking

• SqlResult has now new variants SqlResult::NoData and SqlResult::NeedData.
• SqlResult::into_result_with_trunaction_check has been replaced by SqlResult::into_result_with.
• handles::Environment::driver_connect now returns SqlResult<()> instead of SqlResult<bool>.
• handles::Environment::drivers_buffer_fill now returns SqlResult<()> instead of SqlResult<bool>.
• handles::Environment::drivers_buffer_len now returns SqlResult<(i16, i16)> instead of SqlResult<Option<(i16, i16)>>.
• handles::Environment::data_source_buffer_len now returns SqlResult<(i16, i16)> instead of SqlResult<Option<(i16, i16)>>.
• handles::Connection::execute now returns now returns SqlResult<()> instead of SqlResult<bool>.
• handles::Connection::exec_direct now returns now returns SqlResult<()> instead of SqlResult<bool>.
• handles::Statement::fetch now returns now returns SqlResult<()> instead of SqlResult<bool>.
• handles::Statement::put_binary_batch now returns now returns SqlResult<()> instead of SqlResult<bool>.

Non Breaking

• Reserve an extra word (2 Bytes) when querying Catalog length in order to compensate for an of by one error in PostgreSQL driver when using wide character encoding. The issue would have resulted in the last letter being replaced with \0 in previous versions.

0.49.0

• No breaking changes. Messed up setting version. Should have been 0.48.1 😅.
• Asynchronous execution of one shot queries using Connection::execute_polling.
• Asynchronous bulk fetching of results using RowSetCursorPolling.

0.48.0

• Both Preallocated::row_count and Prepared::row_count now return Option<usize> instead of isize.

0.47.0

• Support getting number of rows affected by INSERT, UPDATE or DELETE.
  • New method Preallocated::row_count
  • New method Prepared::row_count
• TextRowSet::from_max_str_lens now takes IntoIterator<Item=usize> instead of Iterator<Item=usize>.
• trait method RowSetBuffer::bind_to_cursor has been renamed to bind_columns_to_cursor and now takes a StatementRef instead of a Cursor. This change is unlikely to affect user code, as so far I know all downstream crates use the provided RowSetBuffer implementations and do not implement their own versions.

0.46.0

• Minimal support for asynchronous code in the handle module.
• SqlResult has a new variant SqlResult::StillExecuting.
• New function Statement::set_async_enable can be used to enable polling for statements.
• Functions returning Option<SqlResult<_>> now return SqlResult<Option<_>> or SqlResult<bool> in order to allow for top level asynchronous code to build on the same abstraction as the synchronous code.

Since most users would not engage with the unsafe functions of the handle module their code should be unaffected.

0.45.1

• TextRowSet::for_cursor now only performs a faliable allocation, if no upper bound for string length is specified. This has been changed to remidy performance regressions, but still have the faliable allocation in situation there it could likely occurr.

0.45.0

• Add Connection::into_prepared to allow for prepared queries to take ownership of the connection.
• Prepared::describe_param now requires an exclusive reference
• Remove Prebound.

0.44.3

• Add NullableSliceMut::set_cell

0.44.2

• Add Bit::from_bool.

0.44.1

• Add ColumnarBulkInserter::capacity.

0.44.0

• All methods on the ResultSetMetaData trait now require exclusive (&mut) references.
• The trait BorrowMutStatement has been replaced by AsStatementRef.
• ColumnarBulkInserter is now generic over statement ownership.
• It is now possible to create a bulk inserter which just borrows a prepared statement, rather than taking ownership of it. Use Prepared::any_column_inserter to do so. This is useful for dynamically growing the capacity of the bound array parameters. So far users were forced to create a new ColumnarBulkInserter to do so. This is still true, but creating a new ColumnarBulkInserter now longer implies that a new prepared statement has to be created.

0.43.0

• BoundInputSlice, TextColumnSliceMut and BinColumnSliceMut now only track the lifetime of the statement. The do no longer need to track the lifetime of the Connection anymore.

0.42.0

• Removed TextColumn::set_max_len as resize_max_str_len can do everything the former can and does it better.
• ColumnBuffer and TextRowSet can no longer be used to insert data into the database. It can now only be used to receive data. This is due to different invariants which must hold true for input and output buffers. The new ColumnarBulkInserter which can be constructed using into_any_column_inserter or into_text_inserter on a prepared statment closes this safety hole. It also allows for faster insertion into a database in case you transmit the data in several batches, because it does not require the user to rebind the parameter buffers in safe code. This significantly reduces the overhead for each batch.

0.41.0

• Refactored traits for passing parameters to Connection::execute. Should not break user syntax, though.
• Allow using fixed size paramters like i32 directly as output parameter without wrapping them into a Nullable first.
• Renamed TextColumnWriter::rebind into TextColumnWriter::resize_max_str
• Fast insertion into database using TextRowSet without rebinding parameter buffers for individual batches.

0.40.2

• Introduce alias for ColumnarAnyBuffer for ColumnarBuffer<AnyColumnBuffer>.

0.40.1

• Introduce feature iodbc. This features enables all features required for this crate to work with the IODBC driver manager.

  • narrow ODBC function calls
  • Use of odbc_version_3_5 symbols and declaration of ODBC API version 3.0.
  • Linking of upstream odbc-sys vs iodbc as opposed to plain odbc.

• Introduce ColumnDescription::new to allow for intantiating ColumnDescription in test cases without caring about wether SqlChar is u8 or u16.

0.40.0

• The fallibale allocations introduced for buffer_from_description had performance implications. Therfore the fallibale allocations are available to users of this library to try_buffer_from_description. buffer_for_description is now panicing again in case there is not enough memory available to allocate the buffers.

0.39.1

• Fixed an issue introduced in 0.39.1 there the terminating zero at the end of diagnostic messages has not been removed correctly.

0.39.0

• Support for checking for truncation after bulk fetch with: RowSetCursor::fetch_with_truncation_check.

0.38.0

• Add column buffer index to allocation errors to provide more context

0.37.0

• Add fallibale allocation of binary columns.
• Add fallibale allocation of text columns.

0.36.2

• Add AnyColumnViewMut::as_text_view.

0.36.1

• If setting the ODBC version fails with HY092 (Invalid attribute value), the user will now see an error message stating that likely the ODBC driver manager should be updated as it does not seem to support ODBC version 3.80. This gives the user a better starting point to fix the problem than just the diagnostic record.

0.36.0

• Introduce AnyColumnView::Text and AnyColumnView::WText now holds a TextColumnView instead of a TextColumnIt. To get the iterator you can just call iter() on the TextColumnView.
• Introduce AnyColumnView::Binary now holds a BinColumnView instead of a BinColumnIt.
• Add AnyColumnView::as_text_view.
• Add AnyColumnView::as_w_text_view.
• Add AnyColumnView::as_slice.
• Add AnyColumnView::as_nullable_slice.

The TextColumnView has been added in order to allow for faster bindings to arrow. As it gives access to the raw value buffer.

0.35.2

• Allow direct access to raw buffers of NullableSliceMut.

0.35.1

• Allow direct access to raw buffers of NullableSlice.
• ColumnWithIndicator::iter is no longer unsafe
• TextColumn::indicator_at is no longer unsafe
• TextColumn::value_at is no longer unsafe
• BinColumn::value_at is no longer unsafe.
• BinColumn::iter is no longer unsafe.

Thanks to @jorgecarleitao for providing all the input about safety and discovering performance
bottlenecks in the buffer interfaces.

0.35.0

• Better error for Oracle users, then trying to fetch data into 64 Bit Buffers. Added variant OracleOdbcDriverDoesNotSupport64Bit to error type in order to achieve this.

0.34.1

• Introduce feature odbc_version_3_5. Decalaring ODBC version 3.80 is still default behaviour. The flag has mainly been introduced to achieve compatibility with iODBC.

0.34.0

• Introduces narrow feature, allowing to compile this library against narrow ODBC function calls. This is optional now, but may become default behaviour in the future on non-windows platforms (because those typically have UTF-8 locale by default.)
• Removed public higher level methods directly dealing with utf16. Similar functions using unicode remain in the API. This has been done in order better support building against narrow ODBC function calls on some platforms while keeping complexity in check.
• ResultSetMetadata::col_name now returns Result<String, Error>.
• Environment::driver_connect now takes &mut OutputStringBuffer instead of Option<&mut OutputStringBuffer>. Use OutputStringBuffer::empty() instead of None.
• Removed Connection::tables now takes &str instead of Option<&str> arguments. Use "" instead of none
• Removed Environment::connect_utf16.
• Removed Preallocated::execute_utf16.
• Removed Connection::execute_utf16.
• Removed Connection::prepare_utf16.
• Removed OutputStringBuffer::ucstr.
• Removed Connection::fetch_current_catalog.
• Removed Connection::fetch_database_management_system_name.
• updated to odbc-sys 0.21.2

0.33.2

• No changes (Release is currently tied to odbcsv)

0.33.1

• No changes (Release is currently tied to odbcsv)

0.33.0

• Uses Rust Edition 2021
• Mutable references &mut T are no longer implicitly bound as input / output Parameters, but must be wrapped with InOut instead. This has been done to make the decision between Out and InOut explicit, and also help to avoid binding parameters as InOut by accident. Shared references &T are still implicitly bound as input parameters however.
• Renamed Parameter to ParameterRef
• Renamed Parameter::bind_parameter to ParameterRef::bind_to
• Renamed Output to OutputParameter
• Renamed TextColumn::rebind to TextColumn::resize_max_str.
• InputParameter is now longer exported in the top level namespace, but only as parameter::InputParameter.
• ParameterRefCollection::bind_to_statement now takes &mut self rather than self.
• ColumnarRowSet and TextRowSetBuffer are now ColumnarBuffer, which is generic over columns.
• ColumnarRowSet::new is now buffer_from_descriptions.
• Removed SqlResult::log_diagnostics.

0.32.0

• Removed unsafe from trait HasDataType.
• Introduced Prepared::into_statement and CursorImpl::new to enable usecases not covered by safe abstractions yet.

0.31.2

• escape_attribute_value now also escapes + signs. This has been introduced because an authentication vs a PostgreSQL failed, if a password containing + has not been escaped.

0.31.1

• Fix: Some drivers (i.e. SqLite) report the size of column names in characters instead of bytes. This could cause truncation of names provided by colum_names if the supplied buffers were to short. Now the correct provided buffer is always resized to be large enough to hold the entire name, even for these drivers.

0.31.0

In order to avoid exposing the dependency to raw-window-handle in the public interface the way this library opens prompts for connecetion on windows platform has changed:

• Breaking change: Environment::driver_connect now automatically creates a message only window for the user on non-windows platform. DriverCompleteOption is now a simple C-Like enumeration.
• Introduced Environment::driver_connect_with_hwnd to allow power users to provide their own window handles.

0.30.1

• Minor improvements to documentation.

0.30.0

• Allows for fetching result set metadata from Prepared statements.
  • Introduced trait ResultSetMetadata. Which is implemented by both Prepared and Cursor.

0.29.3

• Make StatementConnection public

0.29.2

• Add into_cursor to Connection

0.29.1

• Improved documentation for Connection::columns.

0.29.0

• Removed generic lifetime argument from CursorImpl.
• Add Connection::tables.
• Add Preallocated::tables.

0.28.3

• After failing to get any race conditions with it on either Windows or Linux. Environment::new is now considered to be safe.

0.28.2

• Reexport crate force_send_sync.

0.28.1

• Fix: handles::Environment is now longer declared Sync. Due to the interior mutability regarding error states.
• Fix: A lock has been added to Environment to ensure diagnostics for allocation errors always are handled by the correct thread.

0.28.0

• Add is_empty to NullableSlice and NullableSliceMut.
• Remove Statement::col_data_type. Method is still available via the Cursor trait, though.
• Replace Error with SqlReturn in handles module to not make any decisions about handling diagnostics in that layer.

0.27.3

• Add implementation for ExactSizeIterator for TextColumnIt and BinColumnIt.

0.27.2

• Add NullableSlice::len.

0.27.1

• Fix: DataType::Float is now a struct variant holding a precision value which is either 24 for 32Bit Floating points or 53 for 64Bit Floating point numbers.

0.27.0

• Connection::columns and Preallocated::columns now return a Result<CursorImpl, Error> instead of a Result<Option<CursorImpl>, Error> since columns always returns a Cursor.
• These functions now return a Result<u16, Error> instead of a Result<usize, Error>:
  • Connection::max_catalog_name_len
  • Connection::max_schema_name_len
  • Connection::max_column_name_len
• Renamed OptIt to NullableSlice.
• Renamed OptWriter to NullableSliceMut.
• Added buffer::Item trait to make it more convinient to extract a slice or a NullableSlice from AnyColumnView.

0.26.1

• handles::Nullablility now implements Copy.
• Nullable<T> now implements Copy if T does.

0.26.0

• Add Error::InvalidRowArraySize which is returned instead of Error::Diagnostics if invalid attribute is returned during setting row set array size.

0.25.0

• Error::Diagnostics is now a struct variant and also contains the name of the ODBC function call which caused the error.
• Renamed Error::OdbcApiVersionUnsupported to UnsupportedOdbcApiVersion to achieve consistent word order.
• Fix: Row set array size is now specified in usize rather than u32.
• Fix: Bind type is now specified in usize rather than u32.
• Fix: Parameter set size is now specified in usize rather than u32.

0.24.1

• Fix: Code example blobs
• Fix: An issue intruduced in 0.24.0 there a searched update or delete statement that does not affect any rows at the data source could cause a panic.

0.24.0

• Add Support for sending long data.
• Add variant DataType::LongVarChar.
• Add variant DataType::LongVarBinary.
• Add variant Error::FailedReadingInput.
• trait HasDataType no longer implies trait CData.

0.23.2

• Adds Connection::columns
• Adds Connection::columns_buffer_description

These wrap the ODBC SqlColumns function and represent anoter way to obtain schema information.

Thanks to @grovesNL for the contribution.

0.23.1

• Add support for Connection Pooling
• Add methods for exposing metadata to SQLColumns based queries (Thanks again to @grovesNL).

0.23.0

• Better error message in case ODBC Api version is set on older version of unixODBC.
  • Introduced OdbcApiVersionUnsupported variant to Error type.
  • Changed ODBC state representation in diagnostic to narrow ASCII characters.

0.22.3

• Fix: VarCharBox::null() had been bound as VARCHAR(0). This caused an error with Access and older versions of Microsoft SQL Server ODBC Driver. To fix this VarCharBox::null() is now bound as VARCHAR(1). Thanks to @grovesNL for finding and investigating the issue.

  Analoge issues also fixed for:

  • VarBinarySlice::NULL
  • VarBinaryBox::null()

0.22.2

• Fix: VarCharSlice::NULL had been bound as VARCHAR(0). This caused an error with Access and older versions of Microsoft SQL Server ODBC Driver. To fix this VarCharSlice::NULL is now bound as VARCHAR(1). Thanks to @grovesNL for finding and investigating the issue.
• Support access to raw odbc-sys functionality. This is done to support use cases for which a safe abstraction is not yet available (like streaming large values into the database).
  • Add preallocated::into_statement.
  • Add StatementImpl::into_sys.
  • Add StatementImpl::as_sys.
  • Add Statement::param_data

0.22.1

• Support for passing binary data as parameter.

  • VarBinarySlice - immutable borrowed parameter.
  • VarBinarySliceMut] - mutable borrowed input / output parameter
  • VarBinaryArray - stack allocated owned input / output parameter
  • VarBinaryBox - heap allocated owned input / output parameter
  • &[u8] now implements IntoParameter.
  • Vec<u8> now implements IntoParameter.

• Introduced CursorRow::get_binary to fetch arbitrary large blobs into a Vec<u8>.

• Fix: VarCharArray::new is now placing a terminating zero at the end in case of truncation. As statet in the documentation.

0.22.0

• Add Environment::driver_connect which forwards more functionality from SqlDriverConnect. In particular it is now possible (on windows systems) to complete connection strings via the dialog provided by the ODBC driver and driver manager.
• Error has new variant: Error::AbortedConnectionStringCompletion. Which occurs if the GUI for completion is closed before returning any data.

Thanks to @grovesNL for the contribution.

0.21.0

• Environment::drivers now takes &self instead of &mut self.
• Environment::data_sources now takes &self instead of &mut self.
• Environment::user_data_sources now takes &self instead of &mut self.
• Environment::system_data_sources now takes &self instead of &mut self.
• handles::Environment::drivers_buffer_len now is unsafe and takes &self instead of &mut self.
• handles::Environment::drivers_buffer_fill now is unsafe and takes &self instead of &mut self.
• handles::Environment::data_source_buffer_len now is unsafe and takes &self instead of &mut self.
• handles::Environment::data_source_buffer_fill now is unsafe and takes &self instead of &mut self.

Thanks to @grovesNL for the contribution.

0.20.1

• Test are now also executed against Maria DB
• Add BufferDescription::bytes_per_row.
• Add function escape_attribute_value.

0.20.0

• Required at least Rust 1.51.0 to compile.
• IntoParameter has been implement for String. This is the most likely thing to break your code as calling .into_parameter() would have worked on String before, but would have invoked the implementation on str. Should you want the old behaviour just call my_string.as_str().into_parameter().
• Replaced VarCharRef with VarCharSlice.
• Replaced VarChar32 and VarChar512 with VarCharArray<LENGTH>.
• Replaced VarCharMut with VarCharSliceMut.
• VarChar::copy_from_bytes has been removed. Try constructing with new, from_buffer or using NULL instead.
• VarChar::indicator() now returns an Indicator instead of an isize.
• Replaced VarCharRef::null() with VarCharSlice::NULL.
• Support CString and CStr as input parameters.

0.19.6

• Box<dyn InputParameter> now implements InputParameter.

0.19.5

• Fix: BinColumnWriter::append erroneously inserted a terminating zero at the end of the value.

0.19.4

• Fix: Timestamps buffers are now bound with a default precision of 7 (100ns) in order to avoid an 'invalid precision' error than inserting (occurred with MSSQL). Previously the precision had been 9 (1ns).

0.19.3

Increased support for adaptive buffering during inserts using ColumnarRowSet:

• Add TextColumnWriter::append.
• Add TextColumnWriter::max_len.
• Add TextColumnWriter::rebind.
• Add BinColumnWriter::max_len.
• Add BinColumnWriter::rebind.
• Add BinColumnWriter::append.

Support for directly writing formatted text with fixed length to the underlying buffers.

• Add TextColumn::set_mut.
• Add TextColumnWriter::set_mut.

0.19.2

• Add TextColumnWriter::set_value.

0.19.1

• Fix: Statement::col_data_type now returns DataType::WChar instead of DataType::Other then appropriate.
• Fix: An out of bounds panic with columnar inserts from text buffer, if the last value of the buffer had maximum length in a buffer with the maximum number of rows filled (yes, off by one error).

0.19.0

• Add variant DataType::WChar for fixed sized UTF-16 character data.
• Add method DataType::utf16_len.

0.18.0

• CursorRow::get_text now returns bool with false indicating a NULL field.
• TextColumn is now generic over its character type. The type alias CharColumn for TextColumn<u8> can serve as a drop in replacement for the previous iteration of TextColumn.
• AnyColumnBuffer has a new variant called WText holding a UTF-16 TextColumn<u16> or WCharColumn.
• TextRowSet::for_cursor has a new parameter max_str_len to designate the upper limit for column buffer size.
• Add TextRowSet::indicator_at to allow checking for truncation of strings.

0.17.0

• Introduces Cursor::next_row to fetch data without binding buffers first.
• Allow retrieving arbitrarily large strings from the data source.
• Removed RowSetCursor::unbind due to soundness issue, if fetching continues without any buffers bound.
• Hides internal interface of Cursor trait behind Cursor::stmt.
• RowSetCursor now implements Drop.
• Statement::set_num_rows_fetched now takes an option to allow unbinding a previously set reference.
• Trait OutputParameter has been renamed to Output.
• Trait FixedSizedCType has been renamed to Pod.

0.16.0

• Cursor::fetch is now safe.
• Introduced Connection::preallocate to allow for efficient preallocation of statement handles in interactive clients.
• Renamed Connection::exec_direct_utf16 into Connection::execute_utf16.

0.15.0

• Text column buffers throughout the API now use &[u8] to represent queried strings instead of CStr to account for the possibility of interior nuls.

0.14.0

• Fix: BufferKind::from_data_type now uses display_size rather than column_size then mapping Numeric and Decimal types to Text buffers.
• BufferKind::from_data_type now maps DataType::Time with precision > 0 to BufferKind::Text in order to avoid loosing fractional seconds.
• Renamed DataType::Tinyint to DataType::TinyInt in order to be consistent with DataType::SmallInt.
• Renamed DataType::Bigint to DataType::BigInt in order to be consistent with DataType::SmallInt.
• Removed Application Row Descriptor handle. Likely to be reintroduced later though.
• ColumnarRowSet now has support for variadic binary data types.
• Add support for binary data types.

0.13.2

• Add support for transactions.

0.13.1

• Add support for bulk inserts to ColumnarRowSet.
• New section about inserting values in the guide.

0.13.0

• Add support for output parameters in stored procedures.
  • Input parameters now must be bound by reference.
  • Input/output parameters are bound by mutable reference.
• Input has been renamed to HasDataType.
• Nullable has been renamed to Nullability.
• Update to odbc-sys 0.17.0.

0.12.0

• Add ColumnarRowSet::with_column_indices, to enable binding only some columns of a cursor object.
• Struct DataSourceInfo is now public
• Associated function of Cursor col_type and col_concise_type have been removed in favor of col_data_type.
• Add DataType::utf8_len.
• Fix: TextRowSet::for_cursor now allocates 4 times larger buffers for character data, in order to have enough space for larger utf-8 characters.

0.11.1

• Fix: There has been an integer overflow causing a panic if an ODBC API call generated more than 2 ^ 15 warnings at once.
• Fix: BufferKind::from_data_type now always uses the integer type which fits best, for Decimal and Numeric types with scale 0.

0.11.0

• Removed Cursor::set_row_bind_type.
• Removed Cursor::set_row_array_size.
• Removed Cursor::set_num_rows_fetched.
• RowSetBuffer trait now requires an implementation of bind_type.
• RowSetBuffer trait now requires an implementation of row_array_size.
• RowSetBuffer trait now requires an implementation of mut_num_rows_fetched.
• Add Environment::data_sources.
• Add Environment::user_data_sources.
• Add Environment::system_data_sources.
• TextRowSet now returns CStr instead of &[u8].

0.10.3

• Removed Connection::promote_send_and_sync because it is unclear that diagnostic messages would always be pulled by the correct thread.
• Adds Environment::drivers to list available ODBC drivers.

0.10.2

• Implement IntoParameter for Option<&str>.

0.10.1

• Connection::execute takes now &self instead of &mut self.
• Support promoting Connection to Send or Sync.

0.10.0

• Adds method Prepared::describe_param().
• handle::Statement::bind_parameter changes into handle::Statement::bind_input_parameter, in order take a * const.
• VarCharParam renamed into VarChar.
• TextRowSet::new renamed into TextRowSet::for_cursor.
• trait ParameterCollection has a new method parameter_set_size.
• Parameter trait is now inheriting from Input.
• ColumnBuffer trait is replaced by CDataMut.
• Extended guide with section about parameter binding.

0.9.1

• Environment is now Send and Sync.
• Fix: Executing statements not generating results now does no longer create cursors.
• Fix: Prepared::execute now returns Result<Option<CursorImpl<_>>, Error> instead of Result<CursorImpl<_>, Error>.
• Fix: Cursor::execute is now unsafe.
• Fix: Cursor::fetch is now unsafe.

0.9.0

• Cursor::bind_row_set_buffer is now Cursor::bind_buffer.
• Connection::exec_direct is now Connection::execute.
• Tuple support for passing input parameters.

0.8.0

• Basic support for binding parameters.
• Basic support for prepared queries.
• Due to the new possibility of binding parameters Statement::exec_direct and Statement::execute must now be considered unsafe.
• BindColParameters renamed to BindColArgs.

0.7.0

• Environment::connect now takes &str instead of &U16Str.
• Adds Environment::connect_utf16.

0.6.2

• Logs all diagnostics in case driver returns multiple errors.

0.6.1

• Adds Bit, OptBitColumn, OptI8Column, OptU8Column.

0.6.0

• Extend ColumnDescription enum with Bigint, Tinyint and Bit.

0.5.0

• Introduced DataType into the ColumnDescription struct.
• Renaming: Drops Buffer suffix from OptDateColumn, OptF32Column, OptF64Column, OptFixedSizedColumn, OptI32Column, OptI64Column, OptTimestampColumn.

0.4.0

• Adds buffer implementation for fixed sized types

0.3.1

• Implements FixedCSizedCType for odbc_sys::Time.

0.3.0

• Renamed cursor::col_data_type to cursor::col_type.
• Adds cursor::col_concise_type.

0.2.3

• Adds cursor::col_precision.
• Adds cursor::col_scale.
• Adds cursor::col_name.

0.2.2

• Add trait buffers::FixSizedCType.

0.2.1

• Fix: buffers::TextColumn now correctly adjusts the target length by +1 for terminating zero.

0.2.0

Adds buffers::TextRowSet.

0.1.2

• Adds col_display_size, col_data_type and col_octet_length to Cursor.

0.1.1

• Adds convenience methods, which take UTF-8 in public interface.
• Adds ColumnDescription::could_be_null.
• Adds Cursor::is_unsigned.

0.1.0

Initial release