A newer version of Hazelcast CLC is available.

View latest

Advanced Scripting Guide

This guide shows you how to write Advanced Scripts for CLC.

Introduction

Hazelcast CLC provides scriptability using a Python-like language called Starlark.

The Advanced Scripts have the .star extension. Throughout this guide, we refer to the advanced script as the script.

The code in a script is organized by top-level and optional functions. The main function runs after all top-level statements have completed. For example, the following script prints lines for Top-Level A, Top-Level C, and Main B in that order.

print("Top-Level A")

def main():
    print("Main B")

print("Top-Level C")
python

Top-level code has the following constraints:

  • Variables defined in the top-level become global immutable variables; they cannot be reassigned.

  • Loops are not allowed.

  • Conditional statements are not allowed.

These constraints do not exist in code written in functions. The main function provides a convenient location for your code.

Your First Script

Keeping these points in mind, let’s write our first script. This script prints the total number of entries across all maps. Open your favorite code editor and optionally turn on the Starlark mode, if available, or the Python 3 mode. Save the following code as hello.star:

def main():
    total_entries = 0
    for map_name in object_list("map"):
        total_entries += map_size(name=map_name)
    print("OK There are {} entries.".format(total_entries))
python

This code does the following:

  • Iterates on the Map names in the cluster

  • Retrieves the entry count for the corresponding Maps

  • Sums the entry counts

  • Displays the total number of entries

Before running the script, let’s make sure we have a few Maps in the cluster. Run the following commands in a terminal window to create populated Maps:

clc demo map-setmany -n map1 10
clc demo map-setmany -n map2 20
clc demo map-setmany -n map3 30
bash

You are ready to run the script:

clc script run hello.star
bash

The output is similar to:

OK There are 60 entries.

Data Types

Advanced scripts support various data types. Some of the data types are common to all Starlark scripts. And others are CLC-specific.

Here is a summary of the Starlark-provided types:

just_none = None
a_boolean = True
an_integer = 38
a_float = 42.24
a_string = "Hazelcast"
a_list = [1, "two", 3.33, False]
a_tuple = (1, "two")
a_dictionary = {1: "one", "two": 2}
a_user_function = lambda: "some value"
a_builtin_function = len
python

For further information on these data types, refer to the Starlark documentation.

Data Value

CLC introduces the data type, which contains the data object returned from distributed data structure (DDS) calls, such as map_get. The data object is a serialized object that can be directly passed to another DDS call. It’s crucial to keep these object in their serialized form. Otherwise they might be serialized in a way that cannot be deserialized by CLC, such as Java Serializable objects. By keeping the objects in their original serialized form, CLC can use those objects as keys and values in an opaque way.

Currently, the only way to create data values is by reading one using a DDS call. Here’s some code that does that, save it as data_value.star

map_set("some-key", "some-value")
obj = map_get("some-key")
print(type(obj))
python

Running it:

clc script run data_value.star
bash

Outputs:

data

Once a data object is retrieved, you can check its type or deserialize it if its one of the supported data types. You can use the following methods and functions with a data object:

  • X.type: Returns the serialization identifier.

  • data_type(X): Returns the name of the serializer.

  • decode_data(X): Deserializes the value. Note that CLC cannot deserialize all values. An error will be raised in case the value was not deserialized.

These methods are summarized in the following script, which you can save as data_value2.star:

map_set("some-key", "some-value")
obj = map_get("some-key")
print("String value of the data object     :", obj)
print("Numeric type of the enclosed object :", obj.type)
print("String value of the enclosed object :", data_type(obj))
print("The deserialized object             :", decode_data(obj))
python

Running it:

clc script run data_value2.star
bash

Outputs:

String value of the data object     : data(STRING)
Numeric type of the enclosed object : -11
String value of the enclosed object : STRING
The deserialzed object              : some-value

User Defined Functions

You can define regular functions using the def keyword and anonymous functions using the lambda keyword.

Here’s an example function that calculates the square of the given number. Save it as fun.star:

def sqr(a):
    return a * a

print(sqr(5))
python

Running the script outputs:

25

For further information on functions, refer to the Starlark documentation.

Anonymous functions are useful to define trivial, inline functions. The following example iterates a list of numbers and creates a new list by calling a function with each item. Save the script as anonymous.star:

def main():
    my_list = [1, 2, 3, 5, 10]
    funs = [
        ("identity", lambda x: x),
        ("squared", lambda x: x * x),
        ("10 more", lambda x: x + 10),
    ]
    for label, f in funs:
        new_list = [f(x) for x in my_list]
        print(label, new_list)
python

Running the script outputs:

identity [1, 2, 3, 5, 10]
squared [1, 4, 9, 25, 100]
10 more [11, 12, 13, 15, 20]

For further information on anonymous functions, refer to the Starlark documentation.

Builtin Functions

Map Functions

You can use several Map functions in your advanced scripts. See the script command documentation for the list of the functions available to advance scripts. In this section, you are going to use some of those functions.

The map_set function sets a value for a key in a Map. The key may be any Starlark value or it can be encoded data. The map_get function returns the encoded data for the given key. Save the following script as map1.star.

def main():
    map_name = "num-map"
    for i in range(3):
        map_set(i, "value-{}".format(i), name=map_name)
    print("Map size:", map_size(name=map_name))
    for i in range(3):
        value = map_get(i, name=map_name)
        decoded_value = decode_data(value)
        print("Value:", decoded_value)
python

Running this script outputs:

Map size: 3
Value: value-0
Value: value-1
Value: value-2

Note the value = map_get(i, name=map_name) line. The map_get function returns the encoded data, not the actual value you passed to map_set. This behavior differs from official Hazelcast client libraries, which decode the data and return the decoded value. The reason for this behavior is covered in the Data Value section.

map_key_set returns the list of keys for a member. To make the above example more robust by getting the keys from the cluster instead of manually specifying them. Save the following script as map2.star:

def main():
    map_name = "num-map"
    for i in range(3):
        map_set(i, "value-{}".format(i), name=map_name)
    print("Map size:", map_size(name=map_name))
    for key in map_key_set(name=map_name):
        value = map_get(key, name=map_name)
        decoded_value = decode_data(value)
        print("Value:", decoded_value)
python

You can also use the map_entry_set function to list (key, value) pairs in a map. Let’s modify the script above to use that function. Save the following script as map3.star:

def main():
    map_name = "num-map"
    for i in range(3):
        map_set(i, "value-{}".format(i), name=map_name)
    print("Map size:", map_size(name=map_name))
    for key, value in map_entry_set(name=map_name):
        decoded_value = decode_data(value)
        print("Value:", decoded_value)
python

Utilities

CLC provides the functions below that may be useful in many kinds of scripts.

argv

The argv function returns the arguments passed to the script in a list. The first item (index 0) is the script’s name.

Save the following script as args.star:

def main():
    args = argv()
    print("Script     :", args[0])
    for i, arg in enumerate(args[1:]):
        print("Argument", i + 1, ":", arg)
python
  1. Let’s run it with no arguments:

    clc script run args.star
    bash

    Output:

    Script     : args.star
  2. Let’s run it with 3 arguments:

    clc script run args.star -- foo bar quux
    bash

    Output:

    Script     : args.star
    Argument 1 : foo
    Argument 2 : bar
    Argument 3 : quux
  3. Now let’s try passing a flag to our script. Remember that flags have a dash (-) as a prefix:

    clc script run args.star -- foo --my-flag
    bash

    Output:

    Script     : args.star
    Argument 1 : foo
    Argument 2 : --my-flag

The -- is required after the script path because the flags for the script command can appear anywhere in the command line, even at the end. You must use -- before any flag that must be passed to the script itself to help the argument parser. If no flags are passed to the script, you can omit the double dash, but its use reduces errors.

env

You can access the environment variables using the env command. When run without arguments, it returns the list of (name, value) pairs. Save the following script as env1.star.

def main():
    for name, value in env():
        print(name, "=", value)
python

Running the script outputs all environment variables.

DESKTOP_SESSION = plasma
DISPLAY = :0
EDITOR = /usr/bin/vim
GDK_BACKEND = x11

You can also get the value of a single environment variable, by passing its name to env. Here’s an example:

user_home = env("HOME")
print("My home directory:", user_home)
python

now

now() function returns the current local time in nanoseconds.

You can convert the time to other units using the predefined time units, such as MILLISECOND, SECOND and DAY. For the valid values, see the clc script topic. The following example measures the duration of an object_list operation, and prints the result in various time units.

time = now()
print(time)

You can convert the time from nanoseconds to other time units using the provided built-in values. Here’s an example that shows how to convert nanoseconds to other time units. Save the following script as time.star:

tic = now()
object_list()
toc = now()
took = toc - tic

print("The operation took:")
print("  ", took//NANOSECOND, "nanoseconds")
print("  ", took//MICROSECOND, "microseconds")
print("  ", took//MILLISECOND, "milliseconds")
print("  ", took//SECOND, "seconds")
python

Running it outputs:

The operation took:
   4152911 nanoseconds
   4152 microseconds
   4 milliseconds
   0 seconds

sleep

It is sometimes necessary to wait for a specified period before continuing. sleep command waits for the given time in nanoseconds. You can also specify other time units, such as milliseconds and seconds, as described in the clc script topic. Here’s a script that just waits for two seconds.

print("Falling asleep...")
sleep(2*SECOND)
print("Woke up")
python

Logging

CLC provides several logging functions, which log messages using CLC’s logging mechanism.

  • log_error(STRING_OR_ERROR): Logs a message or an error with the ERROR level.

  • log_warn(STRING): Logs a message with the WARN level.

  • log_info(STRING): Logs a message with the INFO level.

  • log_debug(STRING): Logs a message with the DEBUG level.

If the logging level set with --log.level is higher than the log level of a message, that message is not written to the log. For example, if --log.level error specified then only log_error functions log messages.

The --log.path PATH flag can be used to redirect the log messages to the given file or to STDERR (usually the terminal) if stderr is used.

Here is a sample script to try the interplay of the log messages and logging levels. Save it as logging.star.

log_error("This is an error log")
log_warn("This is is a warn log")
log_info("This is an info log")
log_debug("This is a debug log")
python

First, run it as follows. Note that we set --log.path stderr to see the logs on the screen. Otherwise they would written to the default log file.

clc script run --log.path stderr --log.level error logging.star
bash

As this specifies ERROR level log messages, the output is similar to the following:

2023-11-01T11:24:18+03:00  ERROR  [logging.star:2:10] This is an error log

Let’s try once more with a different log level. This time with DEBUG:

clc script run --log.path stderr --log.level debug logging.star
bash

This time all log messages are displayed and the output is similar to the following:

2023-11-01T21:30:13+03:00  ERROR  [logging.star:2:10] This is an error log
2023-11-01T21:30:13+03:00   WARN  [logging.star:3:9] This is is a warn log
2023-11-01T21:30:13+03:00   INFO  [logging.star:4:9] This is an info log
2023-11-01T21:30:13+03:00  DEBUG  [logging.star:5:10] This is a debug log

Debugging

Debug functions start with double underscores (__). They run only when the debug mode is activated by passing the --debug flag to the script command. They are ignored if the debug mode is not active.

__trace

You can use the trace function to print the script name, line and column of the location whenever it is run. You can also customize the trace message by passing a string to trace. Save the following script as trace.star.

def fun1():
    __trace()
    print("Fun 1")

__trace("Start-up")
fun1()
python

Running the script normally:

clc script run trace.star

Outputs:

Fun 1

That’s because the __trace calls are removed. Try again with the --debug flag:

clc script run --debug trace.star

Outputs:

>> Start-up: trace.star:5:8
>> TRACE: trace.star:2:12
Fun 1

Output

Just like CLC commands, advanced scripts can produce both structured and unnecessary output. The print function produces unnecessary output, which can be suppressed with the --quiet (shortcut -q) flag, and the output command produces structured output. The structured output is written out as a table, but you can change it to other formats using the --format (shortcut -f) flag.

Here’s an example to demonstrate that. Save the script as output1.star:

def main():
    fruits = ["apple", "grapes", "pear"]
    print("There are {} fruits.".format(len(fruits)))
    print("Unnecessary output lines will dissapear when you use -q\n")

    for i, fruit in enumerate(fruits):
        output(Index=i, Fruit=fruit)
python

Let’s first run it normally:

clc script run output1.star
bash

Outputs:

There are 3 fruits.
Unnecessary output lines will dissapear when you use -q

----------------
 Index | Fruit
----------------
     0 | apple
     1 | grapes
     2 | pear
----------------
    OK Returned 3 row(s).

Let’s now use the -q flag to suppress unnecessary output:

clc script run output1.star -q
bash

Outputs:

----------------
 Index | Fruit
----------------
     0 | apple
     1 | grapes
     2 | pear
----------------

As expected, unnecessary output was not produced. Using a different output format also works:

clc script run output1.star -q -f csv
bash

Outputs:

Index,Fruit
0,apple
1,grapes
2,pear

Separating unnecessary and structured output makes it much easier to extract the result using a pipe. This example uses the jq utility: :

clc script run output1.star -q -f json | jq '.Fruit'
bash

Outputs:

"apple"
"grapes"
"pear"

Passing keyword arguments to output for column label and value can sometimes be limiting. This is because only valid Starlark identifiers can be used as the column label. For example, running the following script returns an error:

def main():
    output(1=2)
python

Because of that, CLC also supports passing a dictionary to the output function. The dictionary must be passed as a positional argument and it must be the only argument to output. The script can also be formatted as follows:

def main():
    output({1: 2})
python

The output function handles rows with missing values. Here’s a script to demonstrate that. Save it as output2.star:

def main():
    for i in range(3):
        output({i: "OK"})
python

Running it outputs the following:

--------------
 0  | 1  | 2
--------------
 OK | -  | -
 -  | OK | -
 -  | -  | OK
--------------

Different output formats handle missing values differently. JSON printer just skips the missing values:

clc script run output2.star -f json
bash

Outputs:

{"0":"OK"}
{"1":"OK"}
{"2":"OK"}

Error Handling

Whenever an error occurs, CLC stops the script and prints an error message. This is usually the safest way to handle the error. But there are times when you want to try something that can result in an error, and handle it without giving back the control.

The script run command supports the --ignore-errors flag, which suppresses errors while running the script. You can use the last_error() function to get the error so you can act on it.

Let’s demonstrate that with an example using the decode_data function. Passing a string to decode_data ends with an error; since that function expects an argument of type data. Save the following sctipt as error.star:

def main():
    v = decode_data("foo")
    print(v)
python

And run it:

clc script run error.star
bash

This displays the following error:

 ERROR At error.star:2:19: expected argument 'value' to be data, got string

This time, run the script with the --ignore-errors flag:

clc script run --ignore-errors error.star
bash

The output is now similar to the following:

None

The error in the data_value line is ignored and v is set to None.

Let’s now try to use the last_error function and act on errors. Save the following script as error2.star:

def main():
    v = decode_data("foo")
    if last_error():
        v = "OK Caught the error."
    print(v)
python

Run the script with the --ignore-errors flag:

clc script run --ignore-errors error2.star
bash

This time the output is:

OK Caught the error.

Since you have used the --ignore-errors flag, the program didn’t end with an error in the decode_data line but the error was recorded. Using the last_error function, the existence of an error was checked and v was set to OK Caught the error..

Conclusion

Advanced Scripting is a powerful tool. It supports many use cases, which are not possible using standard CLC scripts

Be sure to check out the script command reference for all commands.