@spec add_connection(
  atom(),
  keyword()
) :: Supervisor.on_start_child()

Adds a new named connection dynamically at runtime.

If the connection does not start — a batch_writer: option that InfluxElixir.Write.BatchWriter refuses, say — the error is returned and nothing is left behind: the name does not resolve and the client's connection state is released. The error never holds the :token.

@spec client() :: module()

Returns the configured client implementation module.

@spec create_bucket(
  InfluxElixir.Client.connection(),
  binary(),
  keyword()
) :: :ok | {:error, term()}

Creates a bucket in InfluxDB v2 (backwards compatibility). opts may carry retention: in seconds and org_id:; see InfluxElixir.Admin.Buckets.create/3.

@spec create_database(
  InfluxElixir.Client.connection(),
  binary(),
  keyword()
) :: :ok | {:error, term()}

Creates a database in InfluxDB v3. opts may carry retention: as a duration string such as "30d"; see InfluxElixir.Admin.Databases.create/3.

@spec create_token(
  InfluxElixir.Client.connection(),
  binary(),
  keyword()
) :: {:ok, map()} | {:error, term()}

Creates a named InfluxDB 3 token; see InfluxElixir.Admin.Tokens.create/3.

@spec delete_bucket(InfluxElixir.Client.connection(), binary()) ::
  :ok | {:error, term()}

Deletes a bucket in InfluxDB v2 (backwards compatibility).

@spec delete_database(InfluxElixir.Client.connection(), binary()) ::
  :ok | {:error, term()}

Deletes a database in InfluxDB v3.

@spec delete_token(InfluxElixir.Client.connection(), binary()) ::
  :ok | {:error, term()}

Deletes an InfluxDB 3 token by name; see InfluxElixir.Admin.Tokens.delete/2.

@spec execute_sql(
  InfluxElixir.Client.connection(),
  binary(),
  keyword()
) :: {:ok, map() | [map()]} | {:error, term()}

Sends a SQL statement to /api/v3/query_sql as it is.

InfluxDB 3 Core refuses every statement that changes data (verified): DELETE, INSERT and UPDATE are {:error, %{status: 400, body: "Error during planning: DML not supported: ..."}}, CREATE and DROP are ... DDL not supported: ..., anything else is a 405. A SELECT returns {:ok, rows} typed like query_sql/3 rows. Remove data with delete_database/2 instead. Client.Local answers the same way; under its :v3_enterprise profile it also runs DELETE FROM m [WHERE ...] and returns {:ok, %{"rows_affected" => n}}.

@spec flush(atom(), timeout()) :: :ok | {:error, :no_batch_writer}

Forces an immediate flush of the batch writer for a connection.

Returns :ok on success or {:error, :no_batch_writer} if no batch writer is configured for the given connection.

The optional timeout is forwarded to BatchWriter.flush/2 and bounds the underlying GenServer.call. See BatchWriter.flush/2 for the default and semantics.

@spec health(InfluxElixir.Client.connection()) :: {:ok, map()} | {:error, term()}

Checks the health of an InfluxDB instance.

@spec list_buckets(InfluxElixir.Client.connection()) ::
  {:ok, [map()]} | {:error, term()}

Lists all buckets in InfluxDB v2 (backwards compatibility).

@spec list_databases(InfluxElixir.Client.connection()) ::
  {:ok, [map()]} | {:error, term()}

Lists all databases in InfluxDB v3.

@spec point(
  String.t(),
  %{required(String.t()) => InfluxElixir.Write.Point.field_value()} | nil,
  keyword()
) :: InfluxElixir.Write.Point.t()

Constructs a new Point struct.

Parameters

• measurement - measurement name
• fields - field key-value pairs (nil is none)
• opts - optional :tags (nil is none) and :timestamp

See InfluxElixir.Write.Point.new/3.

Examples

InfluxElixir.point("cpu", %{"value" => 0.64},
  tags: %{"host" => "server01"}
)

@spec query_flux(InfluxElixir.Client.connection(), binary(), keyword()) ::
  InfluxElixir.Client.query_result()

Executes a Flux query against InfluxDB v2 (backwards compatibility).

@spec query_influxql(
  InfluxElixir.Client.connection(),
  binary(),
  keyword()
) :: InfluxElixir.Client.query_result()

Executes an InfluxQL query against InfluxDB v3.

@spec query_sql(
  InfluxElixir.Client.connection(),
  binary(),
  keyword()
) :: InfluxElixir.Client.query_result()

Executes a SQL query against InfluxDB v3.

Supports transport: :http | :flight option for transport selection (InfluxElixir.Client.HTTP only; Client.Local is in-memory and ignores it).

Common Options

• :database — overrides the connection-level default database.
• :timeout — per-call timeout in milliseconds, else the connection's :timeout, else 30_000. Over HTTP it is Finch's receive timeout; over Flight it is a wall-clock bound on the call and the streaming of the whole result ({:error, :timeout} past it).
• :pool_timeout — per-call Finch pool checkout timeout in milliseconds (HTTP transport; default 5_000). Applies before :timeout.
• :params — map of $name => value placeholder substitutions (HTTP transport only; Flight returns {:error, :params_unsupported_over_flight}).
• :transport — :http (default) or :flight (Arrow Flight gRPC).
• :flight_port — gRPC port for :flight; falls back to the connection's :flight_port, then 443. Pass tls: false for plaintext ports.

Errors

A connection that is neither a keyword list nor a map is {:error, {:invalid_connection, connection}}. A refused query is {:error, %{status: http_status, body: message}} on both transports: Flight's gRPC status is turned into the nearest HTTP one (see InfluxElixir.Flight.Client.query/3), but the two do not always agree on the status (a schema error is 500 over HTTP and 400 over Flight; a missing database is 404 and 500) and Flight's message keeps the engine's prefix (Error while logically planning query: ...).

@spec query_sql_stream(
  InfluxElixir.Client.connection(),
  binary(),
  keyword()
) :: Enumerable.t()

Executes a streaming SQL query, returning a lazy Stream.

Use for large result sets to avoid loading all rows into memory. On the HTTP transport the JSONL response is decoded incrementally with back-pressure, so only one chunk plus a partial line is held in memory regardless of result size.

Because the return value is an Enumerable.t(), errors cannot be returned as an {:error, reason} tuple. Instead, failure classes — a missing database, a non-success HTTP status, or a transport error — are raised as an InfluxElixir.StreamError when the stream is enumerated. This mirrors the {:error, reason} contract of query_sql/3: an outage surfaces as an exception, never as a silent empty result.

Options

• :database — overrides the connection-level default database.
• :params — map of $name => value placeholder substitutions.
• :timeout — per-call receive timeout in milliseconds.

@spec remove_connection(atom()) :: :ok | {:error, term()}

Removes a named connection dynamically at runtime.

The connection's batch writer, if any, writes what it still holds before it stops (see "Shutdown" in InfluxElixir.Write.BatchWriter).

@spec resolve_connection(atom() | InfluxElixir.Client.connection()) ::
  InfluxElixir.Client.connection()

Resolves a connection reference to a config keyword list.

Accepts either an atom name (looked up via Connection.fetch!/1) or a keyword/map config (returned as-is).

Examples

resolve_connection(:trading)
resolve_connection(host: "localhost", token: "t")

@spec stats(atom(), timeout()) :: {:ok, map()} | {:error, :no_batch_writer}

Returns batch writer statistics for a connection.

Returns {:ok, stats_map} or {:error, :no_batch_writer} if no batch writer is configured for the given connection. The optional timeout (default 5_000 ms) bounds the call.

@spec write(
  InfluxElixir.Client.connection(),
  iodata() | InfluxElixir.Write.Point.t() | [InfluxElixir.Write.Point.t()],
  keyword()
) :: InfluxElixir.Client.write_result()

Writes line protocol, or Points, to InfluxDB using the configured client.

Goes through InfluxElixir.Write.Writer, so payloads over 1 KB are gzipped and a [:influx_elixir, :write, ...] telemetry span is emitted. The payload may be:

• line protocol, a binary or iodata (a list of binaries and bytes);
• a InfluxElixir.Write.Point (see point/3), or a list of them. They are encoded with InfluxElixir.Write.LineProtocol.encode/2 for this call's :precision (a DateTime timestamp is written in that unit) and the connection's :api_version. A point no server accepts is that function's {:error, reason} and nothing is sent.

Any other body, or options that are not a keyword list, is the client's error ({:error, {:invalid_body, body}} and {:error, {:invalid_options, opts}} over HTTP), never a raise. To buffer writes and send them in batches, see write_batched/3.

Options

• :database — overrides the connection-level default database.
• :precision — the unit of the timestamps (default nanoseconds); see InfluxElixir.Client.Local for the spellings each server accepts.
• :accept_partial — InfluxDB 3 only. false makes the write all-or-nothing: the first bad line (a parse error or a schema conflict, in line order) rejects the payload, nothing is stored, and the error is {:error, %{status: 400, body: json}} with "line protocol parsing error" and that one line under "data". Default true: good lines are stored and the bad ones reported as a partial write.
• :no_sync — InfluxDB 3 only. true acknowledges the write before it is persisted to the write-ahead log: faster, and a query right after it may not see the points yet (verified).
• :gzip — true compresses every payload, false none; by default only payloads over 1 KB (see InfluxElixir.Write.Writer.write/3).

Examples

point = InfluxElixir.point("cpu", %{"value" => 0.64}, tags: %{"host" => "web01"})
{:ok, :written} = InfluxElixir.write(:default, point)
{:ok, :written} = InfluxElixir.write(:default, [point], precision: :millisecond)

@spec write_batched(
  atom(),
  iodata() | InfluxElixir.Write.Point.t() | [InfluxElixir.Write.Point.t()],
  keyword()
) :: :ok | {:error, term()}

Buffers a write on the batch writer of a connection.

connection_name is a connection with a batch_writer: option (see InfluxElixir.Config). The payload is a InfluxElixir.Write.Point, a list of them, or line protocol (a binary or iodata); Points are encoded for the writer's :precision and the connection's :api_version, and one without a timestamp is stamped with the time of this call. See InfluxElixir.Write.BatchWriter.write/3 for the buffering, the flush triggers and the backpressure.

Returns :ok once the payload is buffered. Other results:

• {:error, :no_batch_writer} - the connection has no writer running, or its writer stopped (the connection was removed) before it took this call
• {:error, :buffer_full} - the writer holds as much as it admits while a batch is being retried
• {:error, reason} - a point that cannot be encoded, a payload that is none of the forms above ({:invalid_payload, payload}), or, with sync: true, the write's own failure

Options

• :sync - true waits until the batch holding the payload has been written (or has failed all its retries) and returns that result, as InfluxElixir.Write.BatchWriter.write_sync/3 does. Default false.
• :timeout - bound on the call in milliseconds (default 60_000; 300_000 with sync: true).

Examples

point = InfluxElixir.point("cpu", %{"value" => 0.64}, tags: %{"host" => "web01"})
:ok = InfluxElixir.write_batched(:default, point)
:ok = InfluxElixir.write_batched(:default, "cpu value=0.64", sync: true)
:ok = InfluxElixir.flush(:default)