# Changelog
## v2.2.0 (2022-10-16)
- Enhancements
- Configuration access for a connection now returns the configured `:otp_app` used (`nil` if none)
- Starting a connection will now issue a warning if the configured `:otp_app` contains no configuration
- Bug fixes
- Read queries now send the proper `Content-Type: application/json` header for InfluxDB v2.x connections without having to explicitly pass `query_language: :flux` as an option ([#76](https://github.com/mneudert/instream/issues/76))
## v2.1.0 (2022-08-02)
- Enhancements
- Log entries for write requests (`%Instream.Log.WriteEntry{}`) now contain the actual HTTP response status returned from the writer
- Support for deleting points using the InfluxDB v2.x "delete" (predicate) API ([#75](https://github.com/mneudert/instream/pull/75))
## v2.0.0 (2022-06-18)
- Enhancements
- Initial query/write support for InfluxDB `2.x` added
- Logging configuration fetched at runtime instead of during compilation
- The HTTP client library can be replaced by providing a custom implementation of the `Instream.HTTPClient` behaviour
- Backwards incompatible changes
- Minimum required Elixir version is now `~> 1.9`
- Passing a host to `ping/1,2`, `status/1,2`, `version/1,2` is now longer supported. Only the hostname configured in the connection module will be used.
- Support for the InfluxDB versions earlier than `1.7.x` is no longer guaranteed
## v1.0.0 (2020-09-06)
- Enhancements
- Documentation is now available inline (`@moduledoc`, ...) with the `README.md` file targeting the repository (development) instead of releases
- Hydrating query results into series structs now parses RFC3339 times into `:nanosecond` timestamps. If you are using `OTP 21.0` or later you will get the full precision, earlier versions are truncated to the `:microsecond` precision supported by `DateTime`!
- Read queries (everything except writes) are now executed directly in the calling process
- Sending asynchronous queries is now limited to write queries
- Writers can modify the worker state during `init` and `terminate` by implementing the callbacks `init_worker/1` and `terminate_worker/1`
- Backwards incompatible changes
- All "administrative query modules" have been removed
- Minimum required Elixir version is now `~> 1.7`
- Support for accessing the system environment for configuration has been removed in favor of initializer functions/modules
- Support for implementing `use Instream.Writer` has been removed in favor of `@behaviour Instream.Writer`
- Support for plural time units (e.g. `:seconds` instead of `:second`) has been removed
- The default JSON decoder has been switched from `:poison` to `:jason`
- The query builder has been removed
- Bug fixes
- `:influxql` has been moved from `:included_applications` to `:applications` to avoid potential problems with release tooling ([#53](https://github.com/mneudert/instream/issues/53))
## v0.22.0 (2019-11-29)
- Enhancements
- Errors from `:hackney.body/1` are now passed to the caller instead of raising a `MatchError`
- Unix socket usage has been documented
## v0.21.0 (2019-05-15)
- Enhancements
- Read queries can now use parameter binding
- The configuration option `:json_encoder` for the library used for JSON encoding has been added
## v0.20.0 (2019-04-19)
- Enhancements
- Queries can be sent as Flux language queries to InfluxDB using `[query_language: :flux]` in the connection or query options
- Soft deprecations (no warnings)
- The query builder has been removed from documentation and will be eventually removed completely. This will be done because of the current limitations of the builder implementation and the InfluxDB move from InfluxQL to Flux as the query language of choice
- Deprecations
- Accessing the system environment by configuring `{:system, var}` or `{:system, var, default}` will now result in a `Logger.info/1` message and will stop working in a future release
- Implementing `use Instream.Writer` has been deprecated in favor of the more explicit `@behaviour Instream.Writer`. Old implementations will trigger a compile time warning until the old macro is removed
- The already (soft) deprecated time units with plural names (e.g. `:seconds`) will now issue `Logger.info/1` messages when used
- Backwards incompatible changes
- Support for the InfluxDB versions earlier than `1.4.x` is no longer guaranteed
## v0.19.0 (2019-01-24)
- Enhancements
- Connections can be declared with compile time configuration defaults that are later overwritten by the application environment
- The library used for JSON decoding can be changed by using the `:json_decoder` configuration
- Soft deprecations (no warnings)
- All "administrative query modules" have been removed from the documentation. They will eventually be removed completely after a proper deprecation phase
- Support for `{:system, "ENV_VARIABLE"}` configuration has been removed from the documentation. It will eventually be removed completely after a proper deprecation phase
- Backwards incompatible changes
- Default configuration has been extended with `host: "localhost"`
- Minimum required Elixir version is now `~> 1.5`
- Missing application configuration will no longer raise during compilation
- Public access to the internal pool module name has been removed
## v0.18.0 (2018-08-11)
- Enhancements
- Configuration can be done on connection (re-) start by setting a `{mod, fun}` tuple for the config key `:init`. This method will be called with the connection module name as the first (and only) parameter and is expected to return `:ok`
- Experimental support to convert plain maps or query results into series structs has been added
- Passwords are automatically redacted from logged queries when using the default logger
- Supporting rules for the Elixir formatter are available
- Bug fixes
- Special characters in url parameter values (like `&` or `?` in a query) are now properly escaped ([#43](https://github.com/mneudert/instream/pull/43))
- Backwards incompatible changes
- Quoting of identifiers/values has been extracted to the package `influxql` ([hex](https://hex.pm/packages/influxql))
## v0.17.1 (2017-12-16)
- Enhancements
- Support for Elixir 1.5 style `child_spec` has been added ([#39](https://github.com/mneudert/instream/pull/39))
## v0.17.0 (2017-12-10)
- Backwards incompatible changes
- Minimum required Elixir version is now `~> 1.3`
## v0.16.0 (2017-09-25)
- Enhancements
- Queries can be executed with an additional option `:pool_timeout` used to wait for an available worker process. This option is independent of other configured/passed timeouts
- System environment configuration can set an optional default value to be used if the environment variable is unset
- Bug fixes
- All queries now use a `GenServer.call/3` timeout of `:infinity`. This allows raising the timeouts for a call above the default of `5_000` ([#38](https://github.com/mneudert/instream/issues/38))
## v0.15.0 (2017-03-16)
- Enhancements
- Every connection method (like `read` or `write`) can set per-call `http_opts` passed on to the `hackney` client
- Querying for data can now return an `{:error, term}` style tuple if the communication with the server failed. For example `{:error, :nxdomain}` if the host that is queried cannot be resolved by `hackney` ([#33](https://github.com/mneudert/instream/issues/33))
- The line writer now accepts a retention policy used for writing ([#34](https://github.com/mneudert/instream/pull/34))
- The line writer will now return an `{:error, term}` style tuple if a problem occurs. For example `{:error, :nxdomain}` if the host that should receive data cannot be resolved by `hackney` ([#33](https://github.com/mneudert/instream/issues/33))
- Timeouts occurring when executing a query are now returned as `{:error, :timeout}` instead of raising ([#33](https://github.com/mneudert/instream/issues/33))
- Backwards incompatible changes
- Series definitions raise upon compilation if the contain a tag and a field with the same name. This is done to prevent the InfluxDB behavior of adding `_1` to such fields when storing them
## v0.14.0 (2016-12-18)
- Enhancements
- Connections can be configured with a default database used when no value is found in method call or series definition
- Fields and tags with the name `:time` will raise after compilation. They are unqueryable and will be dropped by InfluxDB v1.1.0 when trying to write them.
- Queries can return CSV formatted responses from the server when running InfluxDB >= 1.1.0
- Series definitions are now automatically associated with typespecs
- Backwards incompatible changes
- Minimum required Elixir version is now `~> 1.2`
- Minimum required Erlang version is now `~> 18.0`
- Query builder support for `IF NOT EXISTS` has been removed
- Series definitions now require to contain at least one field
- Soft deprecations (no warnings)
- The plural names for precision units have been soft deprecated to be in line with the similar new naming in `Erlang 19.1` and the upcoming `Elixir 1.4.0`. The old types will continue to work but their usage is discouraged. They will be completely deprecated in an upcoming release.
## v0.13.0 (2016-09-11)
- Enhancements
- Configuration has been split into "compile time" and "runtime" parts. This readds the possibility to change configuration values without recompiling the connection modules ([#22](https://github.com/mneudert/instream/pull/22))
- Configuration values can be fetched from the system environment using `{:system, ENV_VAR}`
- Timeouts for queries (individual and connection wide) can be configured ([#21](https://github.com/mneudert/instream/pull/21))
- Query builder supports `LIMIT` and `OFFSET` for queries ([#19](https://github.com/mneudert/instream/pull/19))
- Writing a series struct with one or more empty tags will now properly construct an entry without these tags present
- Soft deprecations (no warnings)
- The precision units `:micro_seconds`, `:milli_seconds` and `:nano_seconds` have been renamed to `:microseconds`, `:milliseconds` and `:nanoseconds` to matches the upcoming `System.time_unit` definition of `Elixir 1.3.0`. Old variables will continue to work but are highly discouraged and will be completely deprecated in an upcoming release.
- Deprecations
- Defining a series without fields is deprecated and will raise in a future version
- Backwards incompatible changes
- Configuring a connection with multiple hosts is no longer supported
- The otp app of a connection is no longer returned when calling `MyConnection.config/0,1`. It is now only available by specifically requesting it via `MyConnection.config([:otp_app])`
- Writer modules now get a map passed with the connection module registered at key `:module` and (if configured) a udp socket for writing at `:udp_socket`
- The `Cluster` namespace of queries has been removed
## v0.12.0 (2016-05-13)
- Enhancements
- Namespace of administrational convenience modules has been changed from `Cluster` to the better matching `Admin`
- Deprecations
- Configuring a connection with multiple hosts has been deprecated. Instead of multiple `:hosts` only a single `:host` is now expected
- The `Cluster` namespace of queries has been changed to `Admin`. For some backwards compatibility the old modules are delegated to the new ones until removed in the next release
- Backwards incompatible changes
- Atoms for defining a series' database are no longer supported
- Atoms for defining a series' measurement are no longer supported
- Support for the JSON protocol has been removed matching its removal in InfluxDB v0.12.0
- Support for the `SHOW SERVERS` statement has been removed matching its removal in InfluxDB v0.13.0
## v0.11.0 (2016-04-14)
- Enhancements
- Hackney options can be configured ([#17](https://github.com/mneudert/instream/pull/17))
- Line writer now prefers a database passed via arguments over the one passed in the datapoint payload
- Logging for requests is available
- Pings can be send to specific servers
- Status requests can be send to a cluster
- Bug fixes
- Series are now compiled with full environment information preventing "/path/to/lib/nofile" to be compiled as the source of `MySeries.Fields` or `MySeries.Tags`
- Backwards incompatible changes
- Runtime connection configuration reading has been removed in favor of compile time inlining
## v0.10.0 (2016-02-27)
- Enhancements
- Default precision (`:rfc3339`) can be explicitly set
- Query builder can construct `CREATE DATABASE` statements
- Query builder can construct `CREATE RETENTION POLICY` statements
- Query builder can construct `DROP DATABASE` statements
- Query builder can construct `DROP RETENTION POLICY` statements
- Query builder can construct `SHOW` statements
- Series database definitions allow anything "evaluating to a string" ([#14](https://github.com/mneudert/instream/pull/14))
- Series measurement definitions allow anything "evaluating to a string" ([#14](https://github.com/mneudert/instream/pull/14))
- Deprecations
- Atoms for defining a series' database are deprecated
- Atoms for defining a series' measurement are deprecated
- Using `if_not_exists` (`CREATE DATABASE`) has been deprecated and will be completely removed once InfluxDB removes it
- Backwards incompatible changes
- `:cluster` query type has been removed in favor of regular `:read` queries
- Retention policy convenience function has been replaced with a more detailed version
## v0.9.0 (2015-12-27)
- Enhancements
- Error messages are returned without leading/trailing whitespace
- Experimental query builder
- Incomplete or missing series definitions raise an error during compilation
- Ping a connection using `Instream.Connection.ping/0`
- Reading queries can be executed directly using `Instream.Connection.query/2`
- Writing query can be executed directly using `Insream.Connection.write/2`
- Deprecations
- Using `Instream.Data.Read` or `Instream.Data.Write` directly is discouraged and will eventually be removed
- Backwards incompatible changes
- The measurement of a series is no longer optional
- The parameter `opts` for `Instream.Data.Read` and `Instream.Data.Write` is no longer optional
- `use`-ing `Instream.Series` without a complete series definition raises during compilation
## v0.8.0 (2015-11-18)
- Enhancements
- Batch writing using `Line` and `UDP` writers ([#10](https://github.com/mneudert/instream/pull/10))
- Fields can be defined with default values
- Tags can be defined with default values
- Writing can be done over UDP using `Instream.Writer.UDP`
## v0.7.0 (2015-10-22)
- Enhancements
- Convenience module for "SHOW DIAGNOSTICS" queries
- Convenience module for "SHOW STATS" queries
- Precision (= epoch) can be passed to read queries
- Precision can be passed to write queries
- Bug fixes
- Pool configuration (size, overflow) is now properly taken from configuration
- Deprecations
- Support of the JSON protocol is deprecated and will be removed once InfluxDB removes it
## v0.6.0 (2015-09-27)
- Enhancements
- `IF NOT EXISTS` can be passed to database creation queries
- Points can be written with explicit timestamps ([#8](https://github.com/mneudert/instream/pull/8))
- Switched default write method to the line protocol
- Tags are optional when writing with the line protocol
- Backwards incompatible changes
- Switched default write method to the line protocol
## v0.5.0 (2015-09-05)
- Enhancements
- Convenience module for `SHOW SERVERS` queries
- Queries can be executed asynchronously
- Support for line protocol
- Backwards incompatible changes
- Write queries return `:ok` instead of `nil`
## v0.4.0 (2015-07-25)
- Enhancements
- Allows using header authentication (basic auth)
- Allows using pre-defined series modules for write queries
- Provides a way to define series as a module (struct)
- Backwards incompatible changes
- Authentication uses headers by default
## v0.3.0 (2015-06-19)
- Enhancements
- Allows managing retention policies
- Dependencies not used in production builds are marked as optional
- Bug fixes
- Authentication is passed using query parameters by default
- Backwards incompatible changes
- Auth configuration is now expected to be a `Keyword.t`
- Queries of type `:host` are now of type `:cluster`
- Remapped `Admin` namespace to `Cluster` to match query types
## v0.2.0 (2015-04-19)
- Enhancements
- Allows accessing raw query results (undecoded binaries) using `Instream.Connection.execute/2`
- Read queries (binaries) can be executed on a database
- Write queries (map data) can be executed on a database
## v0.1.0 (2015-02-23)
- Initial Release
