Writers

q Python

.qsp.write.toAmazonS3[path; isComplete; region]
.qsp.write.toAmazonS3[path; isComplete; region; .qsp.use (!) . flip (
 (`tenant ; tenant);
 (`credentials ; credentials);
 (`onTeardown ; onTeardown))]

Parameters:

name	type	description	default
path	string or symbol or function	The path of an object to write to S3. A function can also be provide that can dynamically set the file path for a given batch of data.	Required
isComplete	:: or function	A binary function that accepts a metadata dictionary as its first argument and data as its second. The return is a boolean to indicate if the upload of the current file should be marked as completed. This will cause any further writes to the given file path to fail. Writes will finish when the first non-partial batch is received for a given path.	Required
region	string	The AWS region to authenticate against.	`us-east-1`

options:

name	type	description	default
tenant	string	The authentication tenant
domain	string	The name of your domain (private config)	""
credentials	string	Name of a credentials secret to mount. See the authentication section below for details	""
onTeardown	symbol	For an in-flight uploads at the time of upload, this indicates the behavior. By default, all in-flight uploads will be left pending (`none`). This is useful for pipelines that need to resume an upload between teardowns (ex. an upgrade). Alternatively, you can either chose to `abort` any pending uploads or `complete` any pending uploads	`none`

For all common arguments, refer to configuring operators

This operator reads an object from an Amazon Web Services S3 bucket as either text or binary.

Teardown Behavior

By default, teardown will leave pending uploads queued so that when the pipeline is restarted the pending uploads are resumed. This behaviour is useful for allowing pipelines to stop and restart for regularly scheduled maintenance (i.e. version upgrades). When an upload remains in a pending state, it will consume resources in Amazon S3 as a partial file upload. By default, Amazon will expunge these partial uploads after 7 days.

If this behavior is undesired, the teardown behavior can be modified to be either abort or complete. If marked as abort, any pending uploads will be terminated. If the upload is resumed, any already processed data will be lost. If marked as complete, any pending uploads will be pushed to S3. However, if the pipeline was to resume, it will start overwriting the previous object with a new object. It is recommended to only use the complete behavior when a path function is provided that will ensure a new file path is used when resuming.

Examples

Uploading a local file:

Note

Permissions

When running this example inside a kxi-sp-worker image, you must first run system "cd /tmp" to avoid encountering permissions errors.

t:([] date: .z.d; name: 1000?10?`3; id: 1000?0ng; units: 1000?100f; price: 1000?1000f);
`:data.csv 0: csv 0: t;
.qsp.run
 .qsp.read.fromFile[`:data.csv]
 .qsp.write.toAmazonS3[`:s3://mybucket/file.csv; ::; "us-west-1"]

Uploading a stream:

The second argument to `write.toAmazonS3` indicates that data should not be uploaded
until an empty batch is received in the stream.
.qsp.run
 .qsp.read.fromCallback[`pub]
 .qsp.write.toAmazonS3[`:s3://mybucket/file.txt; ::; "us-west-1"; {[md;data] 0 ~ count data}]
pub "Lorem ipsum dolor sit amet, consectetur adipiscing elit, "
pub "sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. "
pub "Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat."
At this point, the chunks have been buffered and are waiting to be converted to
an S3 object. Flush the file and write it to S3.
pub ()

sp.write.to_amazon_s3('s3://mybucket/file.csv', region='us-east-1')

Parameters:

name	type	description	default
path	string or symbol or function	The path to write objects to in S3 or a function to change the path for each message that is processed in the stream. If a function is provided, the return must be a string of the file path to write the current batch of data to. When a function is provided, message data is the argument by default, however parameters can be configured to include metadata.	Required
is_complete	:: or function	A binary function that accepts a metadata dictionary as the first argument and data as its second. This function is invoked on each message that is processed and must return a boolean indicating if the current file processing is complete.	Required
region	string	The AWS region to authenticate against.	`us-east-1`

options:

name	type	description	default
tenant	string	The authorization tenant.	""
domain	string	A custom Amazon S3 domain.	""
credentials	string	The secret name for the Amazon S3 credentials. Refer to the authentication section below for more information.	""
on_teardown	symbol	An `AmazonS3Teardown` behavior indicating what to do with any partial uploads that are incomplete when a teardown occurs.	`none`

Returns: A to_amazon_s3 writer, which can be joined to other operators or pipelines.

Examples:

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> sp.run(sp.read.from_callback('publish')
| sp.encode.csv(',')
| sp.write.to_amazon_s3('s3://mybucket/file.csv', region='us-east-1'))
>>> data = pd.DataFrame({
'name': ['a', 'b', 'c'],
'id': [1, 2, 3],
'quantity': [4, 5, 6]
})
>>> kx.q('publish', data)

name,id,quantity
a,1,4
b,2,5
c,3,6

Docker Kubernetes

For Docker based configurations, set the variables in the worker image configuration.

docker-compose:

version: "3.3"
services:
worker:
 image: portal.dl.kx.com/kxi-sp-worker:1.13.0
 environment:
 AWS_ACCESS_KEY_ID: "abcd"
 AWS_SECRET_ACCESS_KEY: "iamasecret"

Docker credentials file for authentication:

To load a custom configuration file into a Docker deployment, mount the credentials file directory into the container and set KXI_SP_CONFIG_PATH to point to the configuration directory.

version: "3.3"
services:
worker:
 image: portal.dl.kx.com/kxi-sp-worker:1.13.0
 volumes:
 - $HOME/.aws/:/config/awscreds
 environment:
 KXI_SP_CONFIG_PATH: "/config"

Next, add the secret name to the S3 reader configuration

.qsp.write.toAmazonS3[`:s3://bucket/hello; ::; "us-east-1"; .qsp.use``credentials!``awscreds]

Now deploying the pipeline will read the credentials from the AWS credentials file. Note that the credentials file name must be credentials.

For Kubernetes deployments, environment variables can be passed via the REST request sent when launching a job.

curl -X POST http://localhost:5000/pipeline/create -d \
 "$(jq -n --arg spec "$(cat spec.q)" \
 '{
        name     : "s3-writer",
        type     : "spec",
        config   : { content: $spec },
        settings : { minWorkers: "1", maxWorkers: "10" },
        env      : { AWS_ACCESS_KEY_ID: "abcd", AWS_SECRET_ACCESS_KEY: "iamasecret" }
    }' | jq -asR .)"

Kubernetes secrets for authentication:

When using a Kubernetes deployment, Kubernetes secrets can be used to install credentials into the worker.

First, create a secret using an AWS credentials file. Take care to ensure the secret is created in the correct namespace.

kubectl create secret generic --from-file credentials=$HOME/.aws/credentials awscreds

Note that the secret must be formatted so that the variable names are lower case and there is a space either side of the equals sign like so:

[default]
aws_access_key_id = abcd1234
aws_secret_access_key = abcd1234

Next, add the secret name to the S3 reader configuration

.qsp.write.toAmazonS3[`:s3://bucket/hello; ::; "us-west-1"; .qsp.use``credentials!``awscreds]

Lastly, when deploying the worker, add a secret to the Kubernetes configuration in the request.

curl -X POST http://localhost:5000/pipeline/create -d \
 "$(jq -n --arg spec "$(cat spec.q)" \
 '{
        name       : "s3-writer",
        type       : "spec",
        config     : { content: $spec },
        settings   : { minWorkers: "1", maxWorkers: "10" },
        kubeConfig : { secrets : ["awscreds"] }
    }' | jq -asR .)"

q Python

.qsp.write.toConsole[]
.qsp.write.toConsole[prefix]
.qsp.write.toConsole[prefix; .qsp.use (!) . flip (
 (`split ; split);
 (`timestamp; timestamp);
 (`qlog ; qlog))]

Parameters:

name	type	description	default
prefix	string	A prefix for output messages.	""

options:

name	type	description	default
split	boolean	Controls how vectors are printed (see below).	0b
timestamp	symbol	Either `local`,`utc`, ` for no timestamp, or `.` to use the UTC time when qlog is false, and no timestamp when it is true.	`.
qlog	boolean	Prints all console logs to a QLog stream.	0b

For all common arguments, refer to configuring operators

This operator adds a console writer to the current stream.

A console writer outputs content to standard out on the current process prefixed with the event timestamp.

Splitting output:

By default, all vectors (lists of the same type) are printed on a single line. By setting the split option to 1b, vectors will be printed on separate lines. Lists of mixed types are always printed on separate lines.

Examples:

Basic writer:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toConsole[];
Callback takes lists of messages, so enlist the single message
publish enlist "hi!"

2021.03.08D19:19:08.171805000 | "hi!"

Write with prefix:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toConsole["INFO: "];
publish enlist "hi!"

INFO: 2021.03.08D19:19:08.171805000 | "hi!"

Write with split vectors:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toConsole[.qsp.use ``split!(::;1b)];
publish enlist 1 2 3 4

2021.09.30D14:01:50.359012847 | 1
2021.09.30D14:01:50.359012847 | 2
2021.09.30D14:01:50.359012847 | 3
2021.09.30D14:01:50.359012847 | 4

Write output to QLog format

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toConsole[.qsp.use ``qlog!(::;1b)]
publish ("hello"; "world")
publish "==="
publish ([] date: .z.p+til 2; open: 2?100f; high: 2?100f; low:2?100f; close: 2?100f)

2021-10-04 11:35:33.189 [console_f08427d1] INFO  "hello"
2021-10-04 11:35:33.189 [console_f08427d1] INFO  "world"
2021-10-04 11:35:33.189 [console_f08427d1] INFO  "==="
2021-10-04 11:36:12.193 [console_f08427d1] INFO  date                          open     high     low      close
2021-10-04 11:36:12.193 [console_f08427d1] INFO  -----------------------------------------------------------------
2021-10-04 11:36:12.194 [console_f08427d1] INFO  2021.10.04D11:36:12.193211900 74.84699 20.68577 85.28164 46.82838
2021-10-04 11:36:12.194 [console_f08427d1] INFO  2021.10.04D11:36:12.193211901 97.91903 36.07874 69.44221 4.251175

sp.write.to_console()

Parameters:

name	type	description	default
prefix	string	A prefix for each line of output.	""

options:

name	type	description	default
split	boolean	Whether a vector should be printed on a single line, or split across multiple lines with one element per line.	0b
timestamp	symbol	A `ConsoleTimestamp` enum value, or the string equivalent.	`.
qlog	boolean	Whether the output should be through qlog.	0b

Returns: A to_console writer, which can be joined to other operators or pipelines.

Examples:

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> sp.run(sp.read.from_callback('publish')
| sp.write.to_console(timestamp = 'none'))
>>> data = pd.DataFrame({
'name': ['a', 'b', 'c'],
'id': [1, 2, 3],
'quantity': [4, 5, 6]
})
>>> kx.q('publish', data)

name id quantity
----------------
a    1  4
b    2  5
c    3  6

q Python

.qsp.write.toKafka[topic]
.qsp.write.toKafka[topic; brokers]
.qsp.write.toKafka[topic; brokers; .qsp.use (!) . flip (
 (`retries ; retries);
 (`retryWait ; retryWait);
 (`options ; options);
 (`topicConfig ; topicConfig);
 (`registry ; registry);
 (`subject ; subject);
 (`autoRegister; autoRegister);
 (`schemaType ; schemaType))]

Parameters:

name	type	description	default
topic	symbol	A Kafka topic name to publish on.	`
brokers	string or string[]	One or more brokers as host:port.	"localhost:9092"

options:

name	type	description	default
retries	long	Max retry attempts for Kafka API calls.	10
retryWait	timespan	Period to wait between retry attempts.	0D00:00:02
options	dictionary	A dictionary of Kafka producer configuration options	()!()
topicConfig	dictionary	A dictionary of Kafka topic configuration options.	()!()
registry	string	Optional URL to a Kafka Schema Registry. When provided, Kafka Schema Registry mode is enabled, allowing for automatic payload encoding.	""
subject	string	A Kafka subject to read schemas from and publish schemas to. If none is provided, uses "-value".	""
autoRegister	boolean	Controls whether or not to generate and publish schemas automatically.	0b
schemaType	string	Schema type to generate, one of: "JSON", "PROTOBUF", "AVRO"	"JSON"

For all common arguments, refer to configuring operators

This operator acts as a Kafka producer, pushing data to a Kafka topic.

A Kafka producer will publish data to a Kafka broker which can then be consumed by any downstream listeners. All data published to Kafka must be encoded as either strings or serialized as bytes. If data reaches the Kafka publish point that is not encoded, it is converted to q IPC serialization representation. Data can also be linked to Kafka Schema Registry to use automatic type conversion.

Beside the above keys, all available Kafka producer configuration options are supported using the options dictionary. This includes properties such as socket.timeout.ms, fetch.message.max.bytes, etc.

To configure the Kafka topic, you should use the topic configuration options.

AVRO is not supported

Encoding messages with AVRO using the Kafka Schema Registry integration is currently not supported.

Reserved keys

metadata.broker.list and producer.id are reserved values and are maintained directly by the Kafka writer. The Kafka writer will error if these options are used.

Localhost broker:

Write data from a function called 'publish' and write to a Kafka topic
'words' with a default broker at 'localhost:9092'.
.qsp.run .qsp.read.fromCallback[`publish] .qsp.write.toKafka[`words]

Both 'Hello' and 'World' are sent to the Kafka topic as independent events.
publish "Hello"
publish "World"

Multiple brokers:

Write data from a function called 'publish' and write to a Kafka topic
'words' to multiple Kafka brokers.
.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toKafka[`words; ("localhost:1234"; "localhost:1235")]

Both 'Hello' and 'World' are sent to the Kafka topic as independent events.
publish "Hello"
publish "World"

Advanced producer configuration:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toKafka[.qsp.use (!) . flip (
 (`topic ; `words);
 (`options ; (!) . flip (
 (`socket.timeout.ms ; 60000);
 (`socket.send.buffer.bytes ; 4194304);
 (`max.in.flight ; 100000)
 ));
 (`topicConfig ; (!) . flip (
 (`compression.type; `gzip);
 (`delivery.timeout.ms; `100000)
 ))
 )];

Publish 'Hello' and 'World' as a single batch publish
publish ("Hello"; "World")

Kafka Schema Registry with existing schema:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toKafka[.qsp.use (!) . flip (
 (`topic ; `words);
 (`brokers ; "localhost:9092");
 (`registry ; "http://localhost:8081"))];

Kafka Schema Registry with generated schema:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toKafka[.qsp.use (!) . flip (
 (`topic ; `words);
 (`brokers ; "localhost:9092");
 (`registry ; "http://localhost:8081");
 (`autoRegister ; 1b);
 (`schemaType ; "JSON"))];

Kafka Schema Registry with custom subject:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toKafka[.qsp.use (!) . flip (
 (`topic ; `words);
 (`brokers ; "localhost:9092");
 (`registry ; "http://localhost:8081");
 (`subject ; "custom-subject"))];

Writing to explicit partitions

Future versions of the Kafka writer will allow writing to explicit partitions/keys.

However, the current Kafka writer will write to unassigned partitions without using an explicit key.

sp.write.to_kafka('topicName')

Reserved keys

metadata.broker.list and producer.id are reserved values and are maintained directly by the Kafka writer. The Kafka writer will error if these options are used.

Kafka configuration options

All available Kafka producer configuration options are supported using the options dictionary. This includes properties such as socket.timeout.ms, fetch.message.max.bytes, etc.

To configure the Kafka topic, you should use the topic configuration options.

Parameters:

name	type	description	default
topic	symbol	The name of a topic.	`
brokers	string or string[]	Brokers identified a `'host:port'` string, or a list of `'host:port'` strings.	"localhost:9092"

options:

name	type	description	default
retries	long	Maximum number of retries that will be attempted for Kafka API calls.	10
retry_wait	timespan	How long to wait between retry attempts.	0D00:00:02
options	dictionary	Dictionary of Kafka publisher configuration options	()!()
topic_config	dictionary	Dictionary of Kafka topic configuration options.	()!()
registry	string	Optional URL to a Kafka Schema Registry. When provided, Kafka Schema Registry mode is enabled, allowing for automatic payload encoding.	""
subject	string	A Kafka subject to read schemas from and publish schemas to. If none is provided, uses "-value".	""
autoRegister	boolean	Controls whether or not to generate and publish schemas automatically.	0b
schemaType	string	Schema type to generate, one of: "JSON", "PROTOBUF", "AVRO".	"JSON"

Returns: A to_kafka writer, which can be joined to other operators or pipelines.

Examples:

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> sp.run(sp.read.from_callback('publish')
| sp.write.to_kafka('topicName'))
>>> data = pd.DataFrame({
'name': ['a', 'b', 'c'],
'id': [1, 2, 3],
'quantity': [4, 5, 6]
})
>>> kx.q('publish', data)

"name,id,quantity"
"a,1,4"
"b,2,5"
"c,3,6"

Advanced Kafka configuration

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> sp.run(sp.read.from_callback('publish')
| sp.write.to_kafka('topicName', options = {
'socket.timeout.ms': 60000,
'socket.send.buffer.bytes': 4194304,
'max.in.flight': 100000
},
topic_config = {
'compression.type': 'gzip',
'delivery.timeout.ms': 100000
}
))
>>> data = pd.DataFrame({
'name': ['a', 'b', 'c'],
'id': [1, 2, 3],
'quantity': [4, 5, 6]
})
>>> kx.q('publish', data)

"name,id,quantity"
"a,1,4"
"b,2,5"
"c,3,6"

q Python

.qsp.v2.write.toDatabase[table; .qsp.use (!) . flip (
 (`database ; database);
 (`deduplicate; deduplicate);
 (`directWrite; directWrite);
 (`overwrite ; overwrite))]
.qsp.v2.write.toDatabase[table; .qsp.use (!) . flip (
 (`target ; target);
 (`deduplicate; deduplicate);
 (`directWrite; directWrite);
 (`overwrite ; overwrite))]

Parameters:

name	type	description	required	default
table	symbol	Table in kdb Insights Database.	Y

options:

name	type	description	required	default
database	symbol or string	kdb Insights Database name. The service endpoint is created by appending '-sm:10001' to the database name.	Required if `target` is null or empty.
target	symbol or string	The target name. Fully qualified service address in the format {host}:{port}.	Required if `database` is null or empty.
retries	long	Maximum number of retries that will be attempted for establishing the initial connection, or reconnecting after the connection has been lost.	N	60
retryWait	timespan	How long to wait between retry attempts.	N	0D00:00:03
deduplicate	boolean	If the outbound stream should drop duplicate messages that may have been created during a failure event. If enabled, the pipeline must produce deterministic data. If `$KXI_ALLOW_NONDETERMINISM` is set, this value will be forced to be false, potentially resulting in duplicate data after failure recover events.	N	1b
directWrite	boolean	Write data directly to the database instead of streaming. Direct table writing is useful for large backfill ingestion where data is mostly historical. When using direct write, the system will use significantly less resources but the data will not be available for query until the entire ingest is complete. At present direct write is only intended for writing to partitioned tables. Attempts to write other table types will error.	N	0b
overwrite	boolean	When using direct write, indicates if data should overwrite existing date partitions. If disabled, data will be merged with existing data.	N	0b

For all common arguments, refer to configuring operators

Examples

Use streaming mode to write to a specified database.

The following examples assume a package called taq has been deployed with a matching database name. The database name resolves to to taq-sm:10001 to match the runtime service.

quotes: flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P-1D; `test; 99.9; 100i; 100.1; 200i)
.qsp.run
.qsp.read.fromExpr["quotes"]
.qsp.v2.write.toDatabase[`quotes; `taq];

Write directly to the database using the specified target.

Set target to be the fully qualified name of the target service. Set directWrite=True to write quotes data directly to the database.

quotes: flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P-1D; `test; 99.9; 100i; 100.1; 200i)
.qsp.run
.qsp.read.fromExpr["quotes"]
.qsp.v2.write.toDatabase[`quotes; `taq-sm:10001; .qsp.use enlist[`directWrite]!enlist 1b];

Using the database and target parameters

These parameters specify the endpoint for the running database. This is used to discover required information. The database parameter is usually used in kdb Insights Enterprise and generally matches the package name being deployed. It derives an endpoint from the package name, e.g. a package called demo would derive an endpoint of demo-sm:10001 which should match the running database. The target parameter is generally used for kdb Insights SDK deployments and should be the fully qualified endpoint of the database service, e.g. demo-sm:10001.

For more information on using the database writer in an end-to-end deployment, see either the kdb Insights SDK example or the Enterprise example.

Operator versions

This operator supports two versions; v1 (default) and v2. You can switch between versions using the api_version parameter.

Unbounded input

The v2 writer provides much improved functionality when dealing with unbounded data streams. However it requires you to manually trigger writedowns. For more detailed information see the Database Ingest page.

sp.write.to_database(table='trade', database='taq', api_version=2)

name	type	description	required	default
table	CharString	Table in kdb Insights Database.	Y	None
database	CharString	kdb Insights Database name. The service address is created by appending '-sm:10001' to the database name.	Required if `target` is None	None
target	CharString	Fully qualified service address in the format {host}:{port}.	Required if `database` is None	None
retries	pykx.LongAtom, pykx.IntAtom, int	Maximum number of retries that will be attempted for establishing the initial connection, or reconnecting after the connection has been lost.	N	None
retry_wait	Timedelta	How long to wait between retry attempts.	N	None
deduplicate	pykx.BooleanAtom, bool	If the outbound stream should drop duplicate messages that may have been created during a failure event. If enabled, the pipeline must produce deterministic data. If `$KXI_ALLOW_NONDETERMINISM` is set, this value will be forced to be false, potentially resulting in duplicate data after failure recover events.	N	None
directWrite	pykx.BooleanAtom, bool	Write data directly to the database instead of streaming. Direct table writing is useful for large backfill ingestion where data is mostly historical. When using direct write, the system will use significantly less resources but the data will not be available for query until the entire ingest is complete. At present direct write is only intended for writing to partitioned tables. Attempts to write other table types will error.	N	None
overwrite	pykx.BooleanAtom, bool	When using direct write, indicates if data should overwrite existing date partitions. If disabled, data will be merged with existing data.	N	False
api_version	int	Used to distinguish version of operator.	Y	None

Returns: A to_database writer, which can be joined to other operators or pipelines.

Examples:

Write to database using v2 and database parameter.

This example assumes a package called taq has been deployed with a matching database name. The database name resolves to to taq-sm:10001 to match the deployed service endpoint.

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> kx.q('quotes:flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P-1D; `test; 99.9; 100i; 100.1; 200i)')
>>> sp.run(sp.read.from_expr('quotes')
| sp.write.to_database(table='trade', database='taq', api_version=2))

Write to database using v2 API and target.

Substitute service-sm:10001 with the fully qualified name of the target service. Set directWrite=True to write directly to the database.

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> kx.q('quotes:flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P-1D; `test; 99.9; 100i; 100.1; 200i)')
>>> sp.run(sp.read.from_expr('quotes')
| sp.write.to_database(table='tableName', target='service-sm:10001', directWrite=True, api_version=2))

Trigger a write-down using v2 and sp.trigger_write.

>>> from kxi import sp
>>> import pykx as kx
>>> kx.q('quotes:flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P-1D; `test; 99.9; 100i; 100.1; 200i)')
>>> sp.run(sp.read.from_expr('quotes')
| sp.write.to_database(table='tableName', target='service-sm:10001', directWrite=True, api_version=2, name='dbWriter'))
>>> sp.trigger_write(["dbWriter"])

For more information on using the database writer in an end-to-end deployment, see either the kdb Insights SDK example or the Enterprise example.

q Python

.qsp.write.toDatabase[table; assembly]
.qsp.write.toDatabase[table; assembly; .qsp.use (!) . flip (
 (`retries ; retries);
 (`retryWait ; retryWait);
 (`timeout ; timeout);
 (`deduplicate; deduplicate);
 (`directWrite; directWrite);
 (`overwrite ; overwrite);
 (`retries ; retries);
 (`retryWait ; retryWait);
 (`mountName ; mountName);
 (`statusTable; statusTable))]

Parameters:

name	type	description	default
table	symbol	A table name. Note that at present if using `directWrite` table name must be a symbol.	Required
assembly	symbol or string	The assembly name	Required, unless the pipeline has been deployed as part of an assembly, in which case the pipeline will default to writing to that assembly.

options:

name	type	description	default
retries	long	Number of times to retry the connection to the storage manager.	`60`
retryWait	timespan	Time to wait between storage manager connection attempts.	`0D00:00:03`
timeout	int	Timeout value	`0Ni` (no timeout)
deduplicate	boolean	If the outbound stream should drop duplicate messages that may have been created during a failure event. If enabled, the pipeline must produce deterministic data. If `$KXI_ALLOW_NONDETERMINISM` is set, this value will be forced to be false, potentially resulting in duplicate data after failure recover events.	`1b`
directWrite	boolean	Bypass RDB & IDB and write to HDB directly for historical data. Direct table writing is useful for large backfill ingestion where data is mostly historical. When using direct write, the system will use significantly less resources but the data will not be available for query until the entire ingest is complete. At present direct write is only intended for writing to partitioned tables. Attempts to write other table types will error.	`0b`
overwrite	boolean	When using direct write, indicates if data should overwrite existing date partitions. If disabled, data will be merged with existing data.	`1b`
mountName	symbol	Name of mount.	`hdb`
statusTable	symbol	If using direct write with an unbounded stream, this is the name of a table to store session statuses.	`

For all common arguments, refer to configuring operators

This operator writes data using kdb Insights Reliable Transport streams, except when doing a direct write.

Automatic Assembly Discovery

If the assembly parameter is not set, and the pipeline is deployed as part of an assembly, the assembly will be automatically set to the assembly containing the pipeline.

kdb Insights Enterprise only

The .qsp.write.toDatabase writer operates only within kdb Insights Enterprise deployments of the Stream Processor at this time. This functionality does not operate within the kdb Insights Stream Processor Microservice. It is dependant on the ability to query deployed assemblies within Kubernetes.

directWrite should not to be used with a reader using the watch option

When using the directWrite option it is not possible to pair this with a reader using the watch option since directWrite relies on a definitive finish point which is not guaranteed when using a watcher.

Basic example publishing streaming data:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toDatabase[`quotes; `assembly];
quotes: flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P; `test; 99.9; 100i; 100.1; 200i)
publish quotes

Using directWrite for historical data:

quotesOld: flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P-1D; `test; 99.9; 100i; 100.1; 200i)
quotesNew: flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P; `test2; 99.9; 100i; 100.1; 200i)
quotes: quotesOld, quotesNew;
.qsp.run
 .qsp.read.fromExpr["quotes"]
 .qsp.write.toDatabase[`quotes; `assembly; .qsp.use enlist[`directWrite]!enlist 1b];
// quotesOld will be written to hdb directly bypassing RDB & IDB.
// quotesNew will be treated as stream data as it will be newer than the HDB purview.

Direct write purviews

When using direct write, data that is older than the purview of the HDB will go directly to disk but will not be available for query until the pipeline completes. All other data will be streamed to the database and be ready for query immediately.

Direct write compatibility

The direct write functionality is intended for batch processing only. As such when using direct write it should be paired with a reader operator with a finish point. Examples would include - read from file, read from cloud storage, read from database or read from expression. Reading from continuous streaming sources is not supported. Some reader operators support partitioning and will automatically scale up the number of workers. Using this with direct write results in undefined behavior. Therefore set the maxWorkers setting to 1 to ensure no scaling.

Direct write overwrite

When using direct write, overwrite is enabled by default. This is to preserve the old behavior of having each date partition in the ingest data overwrite the existing date partition. In a future release, this will be disabled by default making merge the default option. If you intend to continue to use the overwrite behavior, it is recommended that you set it explicitly in your pipeline configuration.

statusTable

statusTable is required when using direct write in an unbounded stream.

When using direct write in an unbounded stream, data which has been received but not yet written down is organized into a session. Triggering a writedown will write an entire session to the database.

If batches come in with no kx_DWSessionID field in the metadata, they will all be staged as a single session, with an automatically generated session ID, prefixed with "kx_default_", that gets updated after each writedown.

If batches come in with the kx_DWSessionID field in the metadata set to a symbol, they will be staged under this session ID and writedown messages must include the kx_DWSessionID field in the metadata, indicating which session ID to write down.

Without a finish event to trigger the writedown, writedowns must be triggered manually by sending a message with the kx_DWWritedown field in its metadata set to 1b. Warnings: - When kx_DWWritedown is present, the data for that batch is ignored. - session IDs cannot be reused. Reusing session ID will result in those sessions being discarded. - This symbol must only contain characters valid in a linux file name. - The overwrite option must be false when using statusTable. - keyed streams are unsupported and will be treated as unkeyed streams.

The status of completed sessions will be written to the status table, which must be defined, and must have the schema ([] sessionID: string; isComplete: boolean). Once the direct write is complete, the status table will be updated to include that session ID, with isComplete set to 1b.

The statusTable field can only be used with direct write.

Operator versions

This operator supports two versions; v1 (default) and v2. You can switch between versions using the api_version parameter.

sp.write.to_database('tableName', 'assembly')

Parameters:

name	type	description	required	default
table	CharString	Table in kdb Insights Database.	Y	None
assembly	CharString	Assembly defining kdb Insights Database. When a pipeline is deployed within an assembly this parameter will default to the name of that assembly.	Y	None
retries	pykx.LongAtom, pykx.IntAtom, int	Maximum number of retries that will be attempted for establishing the initial connection, or reconnecting after the connection has been lost.	N	None
retry_wait	Timedelta	How long to wait between retry attempts.	N	None
timeout	pykx.IntAtom, int	Timeout value.	N	None
deduplicate	pykx.BooleanAtom, bool	If the outbound stream should drop duplicate messages that may have been created during a failure event. If enabled, the pipeline must produce deterministic data. If `$KXI_ALLOW_NONDETERMINISM` is set, this value will be forced to be false, potentially resulting in duplicate data after failure recover events.	N	None
directWrite	pykx.BooleanAtom, bool	Write data directly to the database instead of streaming. Direct table writing is useful for large backfill ingestion where data is mostly historical. When using direct write, the system will use significantly less resources but the data will not be available for query until the entire ingest is complete. At present direct write is only intended for writing to partitioned tables. Attempts to write other table types will error.	N	None
overwrite	pykx.BooleanAtom, bool	When using direct write, indicates if data should overwrite existing date partitions. If disabled, data will be merged with existing data.	N	True
mountName	CharString	Name of mount.	N	None
statusTable	CharString	Required when using direct write with an unbounded stream, this is the name of a table to store session statuses.	N	None
api_version	int	Used to distinguish version of operator.	N	None

Returns: A to_database writer, which can be joined to other operators or pipelines.

Examples:

Write to database using v1 streaming mode.

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> kx.q('quotes:flip `time`ticker`bid`bidSize`ask`askSize!enlist each (.z.P-1D; `test; 99.9; 100i; 100.1; 200i)')
>>> sp.run(sp.read.from_expr('quotes')
| sp.write.to_database('tableName', 'assembly'))

q Python

.qsp.write.toStream[table]
.qsp.write.toStream[table; stream]
.qsp.write.toStream[table; stream; .qsp.use (!) . flip (
 (`prefix ; prefix);
 (`deduplicate; deduplicate))]

Parameters:

name	type	description	default
table	symbol or string	A table name.	Required
stream	symbol or string	The outbound stream	`$RT_PUB_TOPIC`

options:

name	type	description	default
prefix	string	Stream prefix for URL mapping	`$RT_TOPIC_PREFIX`
deduplicate	boolean	If the outbound stream should drop duplicate messages that may have been created during a failure event. If enabled, the pipeline must produce deterministic data. If `$KXI_ALLOW_NONDETERMINISM` is set, this value will be forced to be false, potentially resulting in duplicate data after failure recover events.	`1b`

For all common arguments, refer to configuring operators

This operator writes data using the kdb Insights Stream.

Note

Writing to a kdb Insights Streams facilitates two principal use cases, chaining of pipelines and the ability to write to a database. If explicitly intending to write to a database however it is strongly suggested that you use Database Writer as this includes functionality for the validation of incoming schemas.

sp.write.to_stream('eqOrder')

Stream versus Writer

This functionality will write directly to a reliable transport stream. If wishing to persist to a database you should consider using the to_database writer.

Parameters:

name	type	description	default
table	symbol or string	Name of the table to filter the stream on. By default, no filtering is performed.	Required
stream	symbol or string	Name of stream to publish to. By default, the stream specified by the `$RT_PUB_TOPIC` environment variable is used.	`$RT_PUB_TOPIC`

options:

name	type	description	default
prefix	string	Prefix to add to the hostname for RT cluster. By default, the prefix given by the `$RT_TOPIC_PREFIX` environment variable is used.	`$RT_TOPIC_PREFIX`
assembly (DEPRECATED)		Kdb Insights assembly to write to. By default, no assembly is used.
insights (DEPRECATED)		Whether the stream being published to uses Insights message formats.
deduplicate	boolean	Whether the outbound stream should drop duplicate messages that may have been created during a failure event. If enabled, the pipeline must produce deterministic data. If `$KXI_ALLOW_NONDETERMINISM` is set, this value will be forced to be false, potentially resulting in duplicate data after failure recover events.	`1b`

Returns: A to_stream writer, which can be joined to other operators or pipelines.

Examples:

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> sp.run(sp.read.from_callback('publish')
| sp.write.to_stream('tableName'))
>>> data = pd.DataFrame({
'name': ['a', 'b', 'c'],
'id': [1, 2, 3],
'quantity': [4, 5, 6]
})
>>> kx.q('publish', data)

name id quantity
----------------
a    1  4
b    2  5
c    3  6

q Python

.qsp.write.toKDB[path;prtnCol;table]

Parameters:

name	type	description	default
path	symbol	The database path.	Required
prtnCol	symbol	The name of the partition column.	Required
table	symbol	The name of the table.	Required

For all common arguments, refer to configuring operators

This operator writes an on-disk partition table.

Writing keyed columns

Plugin supports keyed columns but writes to disk as an unkeyed table.

Stream data into a table:

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toKDB[`db;`date;`myTable];
publish ([] date: 2022.01.01 2022.01.02 2022.01.03 2022.01.04 2022.01.05; open: 1 2 3 4 5; close: 6 7 8 9 10);

sp.write.to_kdb('database', 'date', 'trade')

Parameters:

name	type	description	default
path	symbol	The database path.	Required
prtn_col	symbol	The name of the partition column.	Required
table	symbol	The name of the table.	Required

Returns: A to_kdb writer, that writes an on-disk partition table.

Examples:

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> sp.run(sp.read.from_callback('publish')
| sp.write.to_kdb('database', 'id', 'tableName'))
>>> data = pd.DataFrame({
'name': ['a', 'b', 'c'],
'id': [1, 2, 3],
'quantity': [4, 5, 6]
})
>>> kx.q('publish', data)

"name,id,quantity"
"a,1,4"
"b,2,5"
"c,3,6"

q Python

.qsp.write.toProcess[handle]
.qsp.write.toProcess[.qsp.use (!) . flip (
 (`mode ; mode);
 (`target ; target);
 (`async ; async);
 (`queueLength; queueLength);
 (`queueSize ; queueSize);
 (`spread ; spread);
 (`handle ; handle);
 (`retries ; retries);
 (`retryWait ; retryWait))]

Parameters:

name	type	description	default
handle	symbol	Handle of destination process to write to.	`

options:

name	type	description	default
mode	symbol	`function (call) or `table (upsert) (symbol).	`function
target	symbol	Target function name or table name.	""
async	boolean	Whether writing should be async.	1b
queueLength	long	Max async message queue length before flush.	0Wj
queueSize	long	Max number of bytes to buffer before flushing.	1MB
spread	boolean	Treat the pipeline data as a list of arguments in function mode (can not be set with `params` also set).	0b
params	symbol[]	List of any parameters that should appear before the message data in function mode.	()
retries	long	Number of times to retry the connection to the process.	`60`
retryWait	timespan	Time to wait between process connection attempts.	`0D00:00:03`

For all common arguments, refer to configuring operators

This operator writes data to another kdb+ process using IPC.

Dropped connections

During a retry loop, on connection loss and on startup, all processing on the worker is halted until the connection to the output process is re-established or the max number of retries is reached.

Publish to table:

q -p 1234

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.write.toProcess[.qsp.use `handle`mode`target!`::1234`table`output]

q)publish ([] til 2)
q)`::1234 `output
x
-
0
1

Publish to function:

q -p 1234 <<< "upd: {x set y}"

.qsp.run
.qsp.read.fromCallback[`publish]
.qsp.map[{ (`table; x) }]
.qsp.write.toProcess[.qsp.use `handle`target`spread!(`::1234; `upd; 1b)]

q)publish ([] til 2)
q)`::1234 `data
x
-
0
1

Re-connection to output process:

# Quit the process after some time to force a reconnect
q -p 1234

Set number of retries to 2 and wait time between retries to 2 seconds
.qsp.run .qsp.read.fromCallback[`publish]
.qsp.write.toProcess[.qsp.use (!) . flip (
 (`handle ; `::1234);
 (`mode ; `table);
 (`target ; `destTable);
 (`retries ; 2);
 (`retryWait; 0D00:00:02) ]

WARN  [] SPWRITE Connection lost to IPC output process, attempting to reconnect, writer=ipc-91f585df
INFO  [] CONN Closing connection, h=9
INFO  [] SPWRITE Connection re-established, handle=::1234

sp.write.to_process('::8001', 'eqExecution')

Connection

Worker is halted until an IPC connection is established or given up on during a retry loop (on connection loss and on startup), all processing on the Worker is halted until the connection to the output process is re-established or the max number of retries is reached.

Parameters:

name	type	description	default
handle	symbol	The kdb+ connection handle in the form `':host:port'`. Note the leading colon. The host can be omitted to use localhost, e.g. `'::5000'`.	`

options:

name	type	description	default
target	symbol	The name of the function to be called, or the table to be upserted, as defined in the remote kdb+ process.	""
mode	symbol	Whether the target is a function to be called, or a table to be upserted.	`function
sync	boolean	Whether writing should be synchronous or asynchronous.	1b
queue_length	long	Maximum number of async messages permitted in the queue before flushing.	0Wj
queue_size	long	Maximum number of bytes to buffer before flushing the queue. Defaults to one mebibyte.	1MB
spread	boolean	Whether the pipeline data should be treated as a list of arguments for the target function. This option can only be set if in function mode, and cannot be set if `params` is set.	0b
params	symbol[]	A list of parameters that should appear before the message data in function mode.	()
retries	long	Maximum number of retries that will be attempted for establishing the initial connection, or reconnecting after the connection has been lost.	`60`
retry_wait	timespan	How long to wait between retry attempts.	`0D00:00:03`

Returns: A to_process writer, which can be joined to other operators or pipelines.

Examples:

>>> from kxi import sp
>>> import pandas as pd
>>> import pykx as kx
>>> sp.run(sp.read.from_callback('publish')
| sp.write.to_process('::1234', 'tableName'))
>>> data = pd.DataFrame({
'name': ['a', 'b', 'c'],
'id': [1, 2, 3],
'quantity': [4, 5, 6]
})
>>> kx.q('publish', data)

name id quantity
----------------
a    1  4
b    2  5
c    3  6

q Python

.qsp.write.toSubscriber[table; keyCols]

Parameters:

name	type	description	default
table	symbol	The name of the subscription. This is used to subscribe to data emitted from this node.	Required
keyCol	symbol or symbol[]	A key column, or list of key columns. This can be an empty list for unkeyed subscriptions.	Required

For all common arguments, refer to configuring operators

options:

name	type	description	default
publishFrequency	long	The interval at which the node sends data to its subscribers (milliseconds)	500
cacheLimit	long	The max size (number of rows) for an unkeyed snapshot cache. This is not required in keyed mode.	2000

Examples:

Keyed subscription

This example deploys a subscriber node keyed on the sym and exch column, and configured to publish every 500ms. Three rows are received in quick succession for the same key combination. Each one overwrites the previous and if all occur within the same timer period, then only a single row is published to subscribers. If the incoming records were across timer periods, then a single row would be published for each period.

The .qsp.getSnapshotCache API is used to show the state of the snapshot cache. If a new client subscribed, this is what they receive as an initial snapshot.

.qsp.run
 .qsp.read.fromCallback[`publish]
 .qsp.write.toSubscriber[`data; `sym`exch; .qsp.use `name`publishFrequency!(`subscriber; 500)];
publish enlist`sym`exch`time`pos!(`FD;`LSE;2023.06.25D15:00:00.001;100);
publish enlist`sym`exch`time`pos!(`FD;`LSE;2023.06.25D15:00:00.002;100);
publish enlist`sym`exch`time`pos!(`FD;`LSE;2023.06.25D15:00:00.003;100);
.qsp.getSnapshotCache[`subscriber;enlist[`table]!enlist[`data]]

sym exch| time                          pos
--------| ---------------------------------
FD  LSE | 2023.06.25D15:00:00.003000000 100

Unkeyed subscription

This example deploys an unkeyed subscriber node configured to publish every 500ms. Eight rows are received in quick succession and all eight are published to any subscribers. However the snapshot cache only stores the last 5 rows so a new subscription only receives those last 5 in their snapshot.

.qsp.run
 .qsp.read.fromCallback[`publish]
 .qsp.write.toSubscriber[`data; "s"$(); .qsp.use `name`publishFrequency`cacheLimit!(`subscriber; 500; 5)];
publish flip `time`sym`exch`pos!(2023.06.25D15:00:00. + 1000000 * til 8;`FD;`LSE;100);
.qsp.getSnapshotCache[`subscriber;enlist[`table]!enlist[`data]]

time                          sym exch pos
------------------------------------------
2023.06.25D15:00:00.003000000 FD  LSE  100
2023.06.25D15:00:00.004000000 FD  LSE  100
2023.06.25D15:00:00.005000000 FD  LSE  100
2023.06.25D15:00:00.006000000 FD  LSE  100
2023.06.25D15:00:00.007000000 FD  LSE  100

Updating the cache

Occasionally you might want to replace the data in the snapshot cache, either to empty it or replace outdated information. By adding a cacheOperation key to the message metadata, this signals to the subscriber node to replace its cache with the incoming message. You can do this for both keyed and unkeyed nodes.

This example empties the cache when the day changes. It does this by tracking the current date and when it changes, it adds the cacheOperation key to the message metadata with a value of replace. It then publishes an empty table to the subscriber. To empty the cache you can also send a generic null (::).

Apply node

Use an apply to edit the metadata. The map node does not allow you to persist changes to the metadata.

DATE:.z.d;
.qsp.run
 .qsp.read.fromCallback[`publish]
 .qsp.apply[{[op;md;data]
 if[DATE < .z.d;
 DATE::.z.d;
 md[`cacheOperation]:`replace;
 data:0#data;
 ];
 .qsp.push[op;md;data]
 }]
 .qsp.write.toSubscriber[`data; "s"$()]

The previous example cleared the cache, however you may want to replace the cache with a specific dataset. This can be done by feeding the subscriber node from multiple readers. The example below creates a pipeline with two callback readers and joined together with a union to both feed a subscriber node.

The first (configured with the replace callback) is connected to an apply node which sets the cacheOperation metadata to clear the cache with any data it receives.
The second (configured with the upd callback) is joined with the first callback
Both are feeding the subscriber node on the data topic

Two rows are published to the regular publish callback and the snapshot cache contains them both. Then the replace function is called and the snapshot cache shows only that latest row.

Set non-determinism environment variable

To use the union operator you must set the KXI_ALLOW_NONDETERMINISM to "true". This can be done in code or in the pipeline specification.

setenv[`KXI_ALLOW_NONDETERMINISM; "true"]
rep:.qsp.read.fromCallback[`replace]
 .qsp.apply[{[op;md;data]
 md[`cacheOperation]:`replace;
 .qsp.push[op;md;data]
 }]
.qsp.run
 .qsp.read.fromCallback[`publish]
 .qsp.union[rep]
 .qsp.write.toSubscriber[`data; "s"$(); .qsp.use enlist[`name]!enlist `subscriber]
publish enlist`sym`exch`time`pos!(`FD;`LSE;2023.06.25D15:00:00.001;100);
publish enlist`sym`exch`time`pos!(`FD;`LSE;2023.06.25D15:00:00.002;100);
.qsp.getSnapshotCache[`subscriber;enlist[`table]!enlist[`data]]

sym exch time                          pos
------------------------------------------
FD  LSE  2023.06.25D15:00:00.001000000 100
FD  LSE  2023.06.25D15:00:00.002000000 100

replace enlist`sym`exch`time`pos!(`FD;`LSE;2023.06.25D15:00:00.003;100);
.qsp.getSnapshotCache[`subscriber;enlist[`table]!enlist[`data]]

sym exch time                          pos
------------------------------------------
FD  LSE  2023.06.25D15:00:00.003000000 100

name	type	description	default
table	symbol	Name of the subscription. Used to subscribe to data emitted from this node.	Required
key_col	symbol or symbol[]	A key column, or list of key columns. Can be an empty list for unkeyed subscriptions.	Required

name	type	description	default
variable	symbol	The name of the variable to write the output to.	Required
mode	symbol	The writing behavior. Mode can be one of the following options: - `append` outputs the stream into a variable - `overwrite` sets the variable to the last output of the pipeline - `upsert` performs an upsert on table data to the output variable.	`append`

name	type	description	default
variable	symbol	The name of the q variable the data will be written to.	Required
mode	symbol	How to set/update the specified q variable.	`append`

Writers

Amazon S3

Console

Kafka

kdb Insights Database

v2

v1

kdb Insights Stream

KDB

Process

Subscriber

Variable