Tembo's Python Client for PGMQ

Installation

Install with pip from pypi.org:

pip install tembo-pgmq-python

To use the async version, install with the optional dependencies:

pip install tembo-pgmq-python[async]

Dependencies:

• Postgres running the Tembo PGMQ extension.

Usage

Start a Postgres Instance with the Tembo extension installed

docker run -d --name postgres -e POSTGRES_PASSWORD=postgres -p 5432:5432 quay.io/tembo/pg17-pgmq:latest

Using Environment Variables

Set environment variables:

export PG_HOST=127.0.0.1
export PG_PORT=5432
export PG_USERNAME=postgres
export PG_PASSWORD=postgres
export PG_DATABASE=test_db

Initialize a connection to Postgres using environment variables:

from tembo_pgmq_python import PGMQueue, Message

queue = PGMQueue()

Note on the async version

Initialization for the async version requires an explicit call of the initializer:

from tembo_pgmq_python.async_queue import PGMQueue

async def main():
    queue = PGMQueue()
    await queue.init()

Then, the interface is exactly the same as the sync version.

Initialize a connection to Postgres without environment variables

from tembo_pgmq_python import PGMQueue, Message

queue = PGMQueue(
    host="0.0.0.0",
    port="5432",
    username="postgres",
    password="postgres",
    database="postgres"
)

Create a queue

queue.create_queue("my_queue")

Or create a partitioned queue

queue.create_partitioned_queue("my_partitioned_queue", partition_interval=10000)

List all queues

queues = queue.list_queues()
for q in queues:
    print(f"Queue name: {q}")

Send a message

msg_id: int = queue.send("my_queue", {"hello": "world"})

Send a batch of messages

msg_ids: list[int] = queue.send_batch("my_queue", [{"hello": "world"}, {"foo": "bar"}])

Read a message, set it invisible for 30 seconds

read_message: Message = queue.read("my_queue", vt=30)
print(read_message)

Read a batch of messages

read_messages: list[Message] = queue.read_batch("my_queue", vt=30, batch_size=5)
for message in read_messages:
    print(message)

Read messages with polling

The read_with_poll method allows you to repeatedly check for messages in the queue until either a message is found or the specified polling duration is exceeded. This can be useful in scenarios where you want to wait for new messages to arrive without continuously querying the queue in a tight loop.

In the following example, the method will check for up to 5 messages in the queue my_queue, making the messages invisible for 30 seconds (vt), and will poll for a maximum of 5 seconds (max_poll_seconds) with intervals of 100 milliseconds (poll_interval_ms) between checks.

read_messages: list[Message] = queue.read_with_poll(
    "my_queue", vt=30, qty=5, max_poll_seconds=5, poll_interval_ms=100
)
for message in read_messages:
    print(message)

This method will continue polling until it retrieves any messages, with a maximum of (qty) messages in a single poll, or until the max_poll_seconds duration is reached. The poll_interval_ms parameter controls the interval between successive polls, allowing you to avoid hammering the database with continuous queries.

Archive the message after we're done with it

Archived messages are moved to an archive table.

archived: bool = queue.archive("my_queue", read_message.msg_id)

Archive a batch of messages

archived_ids: list[int] = queue.archive_batch("my_queue", [msg_id1, msg_id2])

Delete a message completely

read_message: Message = queue.read("my_queue")
deleted: bool = queue.delete("my_queue", read_message.msg_id)

Delete a batch of messages

deleted_ids: list[int] = queue.delete_batch("my_queue", [msg_id1, msg_id2])

Set the visibility timeout (VT) for a specific message

updated_message: Message = queue.set_vt("my_queue", msg_id, 60)
print(updated_message)

Pop a message, deleting it and reading it in one transaction

popped_message: Message = queue.pop("my_queue")
print(popped_message)

Purge all messages from a queue

purged_count: int = queue.purge("my_queue")
print(f"Purged {purged_count} messages from the queue.")

Detach an archive from a queue

queue.detach_archive("my_queue")

Drop a queue

dropped: bool = queue.drop_queue("my_queue")
print(f"Queue dropped: {dropped}")

Validate the length of a queue name

queue.validate_queue_name("my_queue")

Get queue metrics

The metrics method retrieves various statistics for a specific queue, such as the queue length, the age of the newest and oldest messages, the total number of messages, and the time of the metrics scrape.

metrics = queue.metrics("my_queue")
print(f"Metrics: {metrics}")

Access individual metrics

You can access individual metrics directly from the metrics method's return value:

metrics = queue.metrics("my_queue")
print(f"Queue name: {metrics.queue_name}")
print(f"Queue length: {metrics.queue_length}")
print(f"Newest message age (seconds): {metrics.newest_msg_age_sec}")
print(f"Oldest message age (seconds): {metrics.oldest_msg_age_sec}")
print(f"Total messages: {metrics.total_messages}")
print(f"Scrape time: {metrics.scrape_time}")

Get metrics for all queues

The metrics_all method retrieves metrics for all queues, allowing you to iterate through each queue's metrics.

all_metrics = queue.metrics_all()
for metrics in all_metrics:
    print(f"Queue name: {metrics.queue_name}")
    print(f"Queue length: {metrics.queue_length}")
    print(f"Newest message age (seconds): {metrics.newest_msg_age_sec}")
    print(f"Oldest message age (seconds): {metrics.oldest_msg_age_sec}")
    print(f"Total messages: {metrics.total_messages}")
    print(f"Scrape time: {metrics.scrape_time}")

Optional Logging Configuration

You can enable verbose logging and specify a custom log filename.

queue = PGMQueue(
    host="0.0.0.0",
    port="5432",
    username="postgres",
    password="postgres",
    database="postgres",
    verbose=True,
    log_filename="my_custom_log.log"
)

Using Transactions

To perform multiple operations within a single transaction, use the @transaction decorator from the tembo_pgmq_python.decorators module. This ensures that all operations within the function are executed within the same transaction and are either committed together or rolled back if an error occurs.

First, import the transaction decorator:

from tembo_pgmq_python.decorators import transaction

Example: Transactional Operation

@transaction
def transactional_operation(queue: PGMQueue, conn=None):
    # Perform multiple queue operations within a transaction
    queue.create_queue("transactional_queue", conn=conn)
    queue.send("transactional_queue", {"message": "Hello, World!"}, conn=conn)

To execute the transaction:

try:
    transactional_operation(queue)
except Exception as e:
    print(f"Transaction failed: {e}")

In this example, the transactional_operation function is decorated with @transaction, ensuring all operations inside it are part of a single transaction. If an error occurs, the entire transaction is rolled back automatically.