Imported from melezhik/Sparrow6 (ai/kilocode/SKILL.md). Install upstream with npx skills add melezhik/Sparrow6 --skill kilocode. Copyright stays with the author (MIT).

Sparrow6 Task Check DSL & Raku Regex Expert Skill

This skill provides comprehensive knowledge for working with Sparrow6 Task Check DSL and Raku regular expressions. Use it when you need to write verification rules for structured/unstructured text output or craft Raku regex patterns.

Sparrow6 Task Check DSL

Core Concepts

Purpose:

Regexp-based DSL to verify structured and unstructured text output. Input text is parsed line by line and analyzed on a per-line basis.

Key Terminology:

• TC (Task Check) Language — "The DSL itself, both imperative and declarative"
• Task Check File — DSL code written in a task.check file
• DSL Parser — Program that parses DSL code, parses input data, and verifies input line by line against check expressions
• Search Context — The current scope of input text being examined; can be narrowed by context modifiers

Verification Algorithm:

1. For every check expression: calculate current search context
2. Set check step status as unknown
3. For every line in input text: if line matches expression, mark status succeeded
4. After loop: if status remains unknown, mark failed
5. If all check steps succeeded → overall success; otherwise failure

⚠️ Critical:** A failed check does not terminate the scenario immediately. The parser marks the failure internally and continues. The final result reflects all accumulated failures.

Supported Languages:

Raku, Perl5, Bash, Ruby, Python, PowerShell, PHP.

Sparrow6 provides API functions: matched(), captures(), streams_array(), etc.

DSL Syntax

The TC language consists of these building blocks:

| Building Block | Purpose |

| Comments / blank lines / here documents | Documentation and structure | | Check expressions | Patterns to match input data | | Plain strings | Exact case-sensitive text matching | | Regular expressions | Raku regex patterns | | Assert expressions | Boolean assertions for captured data | | Search context modifiers | sequence, range, within expressions | | Generator expressions | Dynamically generate DSL code | | Code expressions | Execute code for debugging/captures |

Check Expressions

Plain text

Exact, case-sensitive matching.

This text should be here

Regular expressions

Similarly to plain text matching, one may check against regular expressions.

The TC language uses Raku Regular Expressions.

constrains!!!

• you should use Raku regepx in regexp: directives

• you should not enclose regular expression in quotes

• you should escape % with backslash, not with double quotes:

wrong:

regexp: "%"

right:

regexp: \%

WARNING!!!

Raku regular expressions are quite different from Perl regexps, please refer the documentation - https://docs.raku.org/language/regexes to correctly write them

Example:

Input:

2019-04-01
Name: Sparrow6
App Version Number: 0.0.1

DSL:

regexp: \d\d\d\d "-" \d\d "-" \d\d # date in format of YYYY-MM-DD
regexp: "Name:" \s+ \w+ # name
regexp: "App Version Number:" \s+ \d+ "." \d+ "." \d+ # version number

Output:

[task check] stdout match <\d\d\d\d '-' \d\d '-' \d\d> True
[task check] stdout match <'Name:' \s+ \w+> True
[task check] stdout match <'App Version Number:' \s+ \d+ \. \d+ \. \d+> True

Matching multiline strings

Check expressions match text in line by line ( aka single line ) mode, there is no way to match multiline strings ( but see sequence expressions ).

For instance following check fails, even though regular expression is correct:

Input:

A
B

DSL:

regexp: A \n B

But you can match consecutive series of lines using sequence expressions.

Regular expressions pitfalls

• Raku regular expressions are quite different from Perl regexps, please refer the documentation - https://docs.raku.org/language/regexes to correctly write them

• Multiline regular expressions ( unlike in Raku ) are prohibited. Following example of DSL is incorrect:

regexp: \d
\s
\d

This is correct form:

regexp: \d \s \d

If you need to write lengthy regular expression that does not fit your editor screen width, you can use generator: to split the expression by smaller chunks and generate final expression dynamically using one of the supported programming languages

Example:

generator: <<HERE
!raku
say "regexp: ", ("a" x 1000);
HERE

Following is a quick migration guide from Perl to Raku regexps

Spaces

In Raku spaces act differently then in Perl regexps.

1. If you want to check against spaces you either use \s+ or " ".

Examples:

regexp: \s
# next regexp is equivalent to the previous one
regexp: " "

# also following two expressions
# are equivalent
regexp: "Hello World"
regexp: "Hello" \s "World"

2. we use spaces to separate chunks in complex regexp for readability, they never treated as spaces unless they are quoted (see point number 1), for example:

regexp: "Bla" "Bla" \d

will match "BlaBla100" string

Quotes for alpha numeric strings

You don't need quotes for letters and numbers unless there is a space needed to match (see the spaces section):

# this all works without quotes

regexp: Hello 
regexp: ABC
regexp: 1977

Adding quotes for those elements, changes nothing, but readability.

Thus following expressions will do the same as the previous ones:

regexp: "Hello"
regexp: "ABC"
regexp: "1977"

Symbols that require escaping

# symbol has to be escaped by double quotes:

regexp: "#"

Following symbols need to be escaped via backslash if one want to match them:

-
%
~ 
: 
, 
@ 
* 
. 
! 
| 
/ 
\ 
- 
" 
' 
$ 
[ 
] 
( 
) 
+ 
= 
^ 
?

So to match "Hello:"

regexp: Hello\:

Other examples:

Wrong:

foo%password

Right:

foo\%password

Wrong:

foo-password

Right:

foo\-password

Wrong:

password!

Right:

password\!

Digits

note: 4 digits should be in stdout
regexp: \d\d\d\d

Digits and symbols

# match "Hello world"
# followed by date 
# in format MON.DD.YYYY
# where MON - month identified as 3 symbols
# for example JAN.01.2025

regexp: "Hello world" \s <[A..Z>]> ** 3 "." \d\d "." \d\d\d\d 

Mixing quotes and back slashes

When escaping symbols ,.+-:,= choose either double quote or back slash, but not both for readability:

# legid, but not necessary
# to use both \ and " for
# escaping of .
regexp: "\.hello"

# following two expressions
# do the same:

# escaping by back slash
regexp: \.hello

# escaping by quotes
regexp: ".hello"

Zero or more symbols

# 0 or more A
regexp: A*

One or more symbols

# One or more A
regexp: A+

Strictly N symbols

# only four A
regexp: A ** 4

# From two to four A
regexp: A ** 2..4

Any symbol, no symbols at all

# you can use regexp form:
regexp: .*
# or this form:
:any:

Beginning of string

Use ^^ for beginning:

regexp: ^^ "Here we begin"

# or with zero or many
# leading spaces

regexp: ^^ \s* "Look ma at my tabs"

End of string

Use $$ for end of string:

regexp: "wait until the end" \s* $$

See also https://docs.raku.org/language/regexes#Quantifiers

Symbol ranges

Instead of writing [a-z] like in Perl, use this form:

# anything from a to z
regexp: <[a..z]>

# anything from a to z, repeated 3 times
regexp: <[a..z]> ** 3

# anything from a to z, repeated from 1 to 3 times
regexp: <[a..z]> 1 ** 3

See also https://docs.raku.org/language/regexes#Enumerated_character_classes_and_ranges

Alternations

Alternations allows to match from many possible options:

# should be at least something
# out of three colors:
regexp: red || green || blue

To separate alternations from other regexp terms use square brackets:

regexp: "hello, mister" \s [ Red || Green || Brown ] "," \s "would you like a" \s [ tea || coffee ] "?"

See also Alternations

Assert Expressions

Asserts expressions are statements that are true or false.

False assert statement always results in task check failure.

Assert expression should be written in form:

assert: {value} {short description}

value should be one of the following

• 1
• true
• True
• 0
• false
• False

1, true, True are used for true assert statements,

0, false, False for false assert statements.

The second short description parameter of assert expression used for informative purposes, so it's just printed in report

Assert expression examples:

DSL:

assert: 1 this is true
assert: 0 this is false
assert: True this is also true, Raku style 
assert: False this is false, Raku style 
assert: true this is true, python/ruby style 
assert: false this is false, python/ruby style 

Output:

[task check] <this is true> True
[task check] <this is false> False
[task check] <this is also true, Raku style> True
[task check] <this is false, Raku style> False
[task check] <this is true, python/ruby style> True
[task check] <this is false, python/ruby style> False

Asserts are almost always created dynamically with generators. See the next section.

Asserts expressions pitfalls

• Strictly follow expression value format, should be on of the following - 1,true,True,0,false,False If the value is incorrect or empty task check parser will treat it as plan text check expression, following are some incorrect assert expressions examples:

# wrong format of assert value
assert: OK this never works

# empty assert value:
assert: ok

• When generating assert expressions via generators follow programming languages documenatation on how to convert language native types into assert expression value format

Check expressions pitfalls

If a check expression check fails it does not terminate the whole scenario straight away, the failed checks is marked in internal state and will result in whole scenario failure when scenario is finished.

So it's wrong to assume if we reach the end of task.check there is no check failures

In other words even though a check is failed parser mark it state as failed and goes to the next check

Matched data and capturing

Don't do that inside code: or generator, it wrong!

generator: <<HERE
!raku
my $input = slurp();
my @lines = $input.lines;
HERE

Instead use capture(), captures() or matched() functions, see later:

generator: <<HERE
!raku
for captures() -> $c {
    my $name = $c[0];
    my $v =  $c[1];
}
HERE

• Parser matches all the lines of input text against check expression

• If at least one line matches a check expression - check step is considered as successful

• Matched lines are captured and accumulated inside matched array and available through matched() function wich returns list of matched strings:

generator: <<HERE
!raku
for matched() -> $m {
}
HERE

• If regular expressions use capturing, the respected captures are kept inside captures array and available through capture() and captures() functions, say we have input in following format:

key1 = value1
key2 = value2

Following code will capture all key/value pairs:

regexp: (\w+) \s+  '=' \s+ (\w+)
generator: <<HERE
!raku
for captures() -> $c {
    my $key = $c[0];
    my $value = $c[1];
}
HERE

capture() function returns the very first element of captures() list and works well when you there only one matched occurrence

Layer

Layer represents all captures for a single check step, if checks are applied within search modifiers (between:, within:, begin:) there maybe more then one layer - one layer per group, see also streams

Examples:

Input:

1 - for one
2 - for two
3 - for three

DSL:

regexp: (\d+) \s+ '-' \s+ for \s+ (\w+)

code: <<CODE
!perl

use Data::Dumper; 

print Dumper(captures());

CODE

Output:

[task check] stdout match <(\d+) \s+ '-' \s+ for \s+ (\w+)> True
[task check] $VAR1 = [
[task check]           [
[task check]             '1',
[task check]             'one'
[task check]           ],
[task check]           [
[task check]             '2',
[task check]             'two'
[task check]           ],
[task check]           [
[task check]             '3',
[task check]             'three'
[task check]           ]
[task check]         ];

Likewise matched() function return array of matched lines:

For the same input:

DSL:

regexp: \d

code: <<CODE
!perl

use Data::Dumper; 

print Dumper(matched());

CODE

Output:

[task check] stdout match <\d> True
[task check] $VAR1 = [
[task check]           '1 one',
[task check]           '2 two',
[task check]           '3 three'
[task check]         ];

The difference between matched and captures() is that matched() returns array of matched lines, while captures() returns array of captured variables for every matched line ( that means array of arrays ), if regular expression has capturing;

In case there is no capturing in regular expression or plain strings check used captures() is equivalent of matched()

capture() function returns first element of the captures() array.

You can do more with captured data then just dumping it out. One of the common cases is to compare captures with certain values, using asserts. Add this code to the previous example:

DSL:

generator: <<CODE
!perl
  print "assert: ", captures()->[0]->[0] == 1 ? 1 : 0, " captures0,0 == 1\n";
  print "assert: ", captures()->[0]->[1] eq 'one' ? 1 : 0, " captures0,1 == one\n";
  
  print "assert: ", captures()->[1]->[0] == 2 ? 1 : 0, " captures1,0 == 2\n";
  print "assert: ", captures()->[1]->[1] eq 'two' ? 1 : 0, " captures1,1 == two\n";
  
  print "assert: ", captures()->[2]->[0] == 3 ? 1 : 0, " captures2,0 == 3\n";
  print "assert: ", captures()->[2]->[1] eq 'three' ? 1 : 0, " captures2,1 == three\n";
  
  print "assert: ", capture()->[0] == 1 ? 1 : 0, " capture0 == 1\n";
  print "assert: ", capture()->[1] eq 'one' ? 1 : 0, " capture1 == one\n";


CODE

And run check:

[task check] <captures0,0 == 1> True
[task check] <captures0,1 == one> True
[task check] <captures1,0 == 2> True
[task check] <captures1,1 == two> True
[task check] <captures2,0 == 3> True
[task check] <captures2,1 == three> True
[task check] <capture0 == 1> True
[task check] <capture1 == one> True

That allows you to write tests in Perl5 Test::More cmp_ok way.

matched(), captures() and capture() are defined for all DSL compatible languages.

Matched data and captures are reset for every check step, meaning that the related functions return data for the last check step.

But also see streams, that accumulated captures within search context.

Capture(), captures(), matched() pitfalls

Captured or matched data is discarded with the next check expression.

In other words capture(), captures(), matched():

All those functions only store data matched or captured for the last check expression

They don't accumulate data for the previous checks.

Consider this example:

Input:

1 hello
2 world

DSL:

regexp: \d \s hello
regexp: \d \s world

code<< CODE
!raku
for matches()<> -> $l {
   say $l
}

This will only print 2 world as this is a line where the last check expression matched

If you want to accumulate matched or captured data consider using within:, between: or begin: together with streams(), streams_array() functions. Streams always accumulate matches data when used together with mentioned search scope modifiers

So as a rule use code: or generator: expression after every check expression to handle captured, matched data:

regexp: \d
code: <<CODE
!raku
# handle matched data
CODE

regexp: \w
code: <<CODE
!raku
# handle matched data
CODE

Code & Generator Expressions

Code expressions:

Execute code for debugging or data extraction.

Start with code: marker followed by <<HEREDOCMARKER, the first line of code: block should be language identifier - !language

code: <<HERE
!language
# code here
HERE

Example for Raku:

code: <<HERE !raku say "hello"; HERE

Code expressions

Code expressions are just pieces of generic language code gets inlined and executed during verification process.

Example:

Input:

Hello

DSL:

Hello
code: <<CODE
!perl
print "hi there!\n";
CODE
Hello

Output:

[task check] stdout match <Hello> True
[task check] hi there!
[task check] stdout match <Hello> True

Code expressions have no impact on verification process and could be used for debugging.

Code expressions should start with code: marker, followed by <<HERE doc marker that defines the start of code block.

The very first line of code block should contain language identificator in form of !language to define programming language to be used to execute code block, for instance:

!python

More code expression examples:

raku

code:  <<HERE
!raku
say 'hi there!'
HERE

bash

code:  <<HERE
!bash
echo 'hi there!'
HERE

ruby

code: <<CODE
!ruby
puts 'hi there!'
CODE

Sparrow6 task API is available inside code expressions:

code: <<CODE
!perl
my $foo = config()->{foo};
CODE

Read Sparrow6 Development Guide on task API.

See also generators section on how dynamically create check expressions using various programming languages.

Code expressions could use all the functions provided by Sparrow Task SDK - https://github.com/melezhik/Sparrow6/blob/master/documentation/development.md

For example to pass input parameter to code: block use config() function:

code: CODE
!python
from sparrow6lib import *
cfg = config()
a = cfg["a"]
CODE

Code expressions pitfalls

• Don't use slurp() inside generator: , use shared state or captures(), matched() instead

• Don't use code expressions to print another DSL expressions ( for instance - regexp:, generator:, code:, within:, begin:, between: , etc ) , this does not make a sence as DSL parser does not treat output from code: expressions as check rules, it just dumps the output as is. Instead use generator: to generate new check rules or DSL expressions

• Always set language identifier on the very first line of code block, or else default language (Perl) will be applied which is not probably what you want

• For code expressions written on Python if Sparrow SDK functions are used ( matched(), captures(), capture(), streams-array(), config(), os(), get_state(), update_state(), etc ) you have to import those functions from sparrow6lib to make them available inside code expression block:

code: <<CODE
!python
from sparrow6lib import *
c = capture()
conf = config()
CODE

Generator expressions:

Dynamically generate DSL rules.

Output printed to stdout is parsed as new DSL code.

Start with generator: marker followed by <<HEREDOCMARKER, the first line of code: block should be language identifier - !language

generator: <<HERE
!language
# code that prints DSL rules
HERE

Example for Raku:

code: <<HERE !raku say "regexp: hello"; HERE

Available API functions (all languages):

• matched() -- returns array of matched lines for the last check expression
• captures() -- returns array of arrays of captured variables per matched line
• capture() -- returns first element of captures() array
• streams_array() -- accumulates matches within search context modifiers

Critical:

matched(), captures(), and capture() only store data for the last check expression. They do not accumulate across checks unless used with within:, between:, or begin: together with streams()/streams_array().

Generators

• Generators is the way to _generate check expressions on the fly

• Generator expressions like code expressions are just pieces of executable code

• Everything that generator code print to stdout is treated as new DSL code and parsed by DSL parser

• So new DSL terms are passed back to DSL parser and executed immediately

Generator expressions should start with generator: marker, followed by <<HERE doc marker that defines the start of generator block.

The very first line of generator block should contain language identificator in form of !language to define programming language to be used to execute generator block, for instance:

!python

The end of generator block should be closed by HERE doc marker.

More generator examples:

Input text:

HELLO

DSL:

generator: <<CODE
!perl
print join "\n", ('H', 'E', 'L', 'O');
CODE

Output:

[task check] stdout match <H> True
[task check] stdout match <E> True
[task check] stdout match <L> True
[task check] stdout match <O> True

Here are examples for other languages:

Input text:

Say
Hello
First

This generator creates 3 new check expressions:

Bash:

generator: <<CODE
!bash
  echo Say
  echo Hello
  echo 'regexp: Hello || Again'
CODE

Ruby:

generator: <<CODE
!ruby
  puts 'Say'
  puts 'Hello'
  puts 'regexp: Hello || Again'
CODE

Output:

[task check] stdout match <Say> True
[task check] stdout match <Hello> True
[task check] stdout match <Hello || Again> True

Here is more complicated example of using Perl language.

Input:

foo value
bar value

DSL:

# this generator creates
# comments
# and plain string 
# check expressions

generator: <<CODE
!perl
 my %d = ( 'foo' => 'foo value', 'bar' => 'bar value' );

 print join "\n", map { ( "# $_ ", $d{$_} ) } keys %d;
  
CODE

Output:

[task check] stdout match <bar value> True
[task check] stdout match <foo value> True

Generators could produce not only check expressions but code expressions and another generators.

Generators are often used to create an asserts from captures.

This is short example for Ruby language:

number: (\d+)

generator: <<CODE
!ruby
    puts "assert: #{capture()[0] == 10}, you've got 10!"  
CODE

Generator expressions could use all the functions provided by Sparrow Task SDK - https://github.com/melezhik/Sparrow6/blob/master/documentation/development.md

For example to pass input parameter to generator use config() function:

generator: CODE
!python
from sparrow6lib import *
cfg = config()
a = cfg["a"]
CODE

Generators pitfalls

• Don't use slurp() inside generator: , use shared state or captures(), matched() instead

• Always set language identificator on the very first line of generator block, or else default language (Perl) will be applied which is not probably what you want

• For generator expressions written on Python if Sparrow SDK functions are used (matched(), captures(), capture(), streams-array(), config(), os(), get_state(), update_state(), etc you have to import those functions from sparrow6lib to make them available inside generator expression block:

• Don't print anything starting with # from generator: code, use note:, as anything started with # will be ignored by DSL parser

Example, don't do that:

generator <<RAKU
!raku
say "# debug message";
RAKU

instead do this:

generator <<RAKU
!raku
say "note: debug messages";
RAKU

generator: <<CODE
!python
from sparrow6lib import *
c = capture()
conf = config()
CODE

Sharing data between code or generator blocks

To share captured data across code: or generator: blocks use functions update_state(), get_state() provided by Sparrow Task SDK:

regexp: ID \: \s+ (\d+)

code: <<CODE
!raku
if matched() {
   update_state(%(
     id => capture()[0]
   ));
}
CODE

regexp: Name \: \s+ (\S+)

code: <<CODE
!raku
if matched() {
   my $id = get_state()<id>;
   my $name = capture()[0];
   say "{$name}'s ID is {$id}";
}
CODE

get_state(), update_state() pitfalls

SDK function get_state() should not accept any arguments it return state.
update_state(state) SDK function should accept state as argument. Please take this into account,when saving state inside multiple code: generator: always get current state first via get_state(), and add new fields/objects to this current state, then save whole updated state:

code: <<HERE
!raku
my $s = get_state();

$s<new-field> = %(
    data => [1,2,3]
);

update_state($s);
HERE

Search Context Modifiers

Sequence expressions:

Match consecutive lines in order. Must be enclosed in begin:/end: markers:

begin:
  line1
  line2
  line3
end:

📇️ You cannot skip blank lines in sequences. Sequence blocks cannot be nested.

Range expressions:

Match lines between two delimiters using between: {start} {end}:

between: {foo} {bar}
  regexp: pattern
end:

⚠️ Ranges verify lines are within bounds, not necessarily consecutive. Nesting ranges is not supported.

Within expressions:

Narrow search context to lines matching a pattern. Requires end: to restore context.

Pitfalls & Best Practices

1. Always use end: to close context modifiers. Forgetting end: leaves the parser in modified context indefinitely.

2. Raku regex spaces: In DSL regex, spaces separate tokens for readability. Quote them or use \s to match literal spaces.

3. Don't use code: to print DSL expressions. Use generator: instead; code: output is not parsed as rules.

4. Failures accumulate. One failed check doesn't stop execution; all checks run.

5. Captures reset per check. Use within:/begin: with streams() for accumulated data.

6. Assert format strictness. Invalid assert values become plain text checks, not assertions.

Negations

Negations are reverse matches, they work the same way as regular expressions but with reverse logic - if text matches - check fails.

Here are some examples:

Input:

OK
Hello
done
A B C D

DSL:

note: negations in blocks

begin:
regexp: OK
!regexp: ok
Hello
done
end:

note: negations for single strings
!regexp: Ok

note: negations inside ranges
note: between: {OK} {done}
between: {OK} {done}
  !regexp: HELLO
  Hello
end:

within: "A B"
!regexp: F
C
end:

output:

[task check]
# negations in blocks
stdout match (s) <OK> True
stdout match (s) <!ok> True
stdout match (s) <Hello> True
stdout match (s) <done> True
# negations for single strings
stdout match <!Ok> True
# negations inside ranges
# between: {OK} {done}
stdout match (r) <!HELLO> True
stdout match (r) <Hello> True
stdout match (w) <!F> True
stdout match (w) <C> True

Search by line numbers AKA SLN search

Sometime it makes a sense to search by line number of text output.

Task check provides SLN ( Search by Line ) exprrssions for that. SLN expression is in form:

colon line number colon

Where line number is a number of line of input text.

There should not be any spaces between line number and colons.

Examples:

# the very first line
# line number 0
# internally input starts with 0
# index

:0:


# line number 99

:99:

Full example:

Input:

OK
HELLO
OK
DONE
BYE!

DSL:

OK

code: <<RAKU
!raku
use Data::Dump;
for captures-full() -> $s {
  say $s;
}
RAKU


begin:
HELLO
!regexp: OK2
:2:
!regexp: OK3
:3:
BYE!
end:

Output:

[task stdout]
13:02:02 :: OK
13:02:02 :: HELLO
13:02:02 :: OK
13:02:02 :: DONE
13:02:02 :: BYE!
[task check]
stdout match <OK> True
# [{data => [OK], index => 0, stream-id => (Any)} {data => [OK], index => 2, stream-id => (Any)}]
stdout match (s) <HELLO> True
stdout match (s) <!OK2> True
stdout match (s) <:2:> True
stdout match (s) <!OK3> True
stdout match (s) <:3:> True
stdout match (s) <BYE!> True

The search starts with "OK" lines check that find two lines. The first is at the index 0, and the second at the index 2.

Next sequential search begins with "HELLO" line (the one and only here), then references to the second "OK" line (at the index number 2), and then to the line with index number 3 (which happens to be a "DONE" line), that all results in effective search of HELLO->OK->DONE->BYE! sequential block.

Additional negation checks make it sure that there is no "OK2" line after the first "HELLO" line, and there is no "OK3" line after the second "OK" line.

You can use SLN in ranges, sequential blocks and within expressions:

note: search everything between 1st and 3d lines
between: { :1: } { :3: }
    regexp: .*
end:

note: start sequence from 2nd line
begin:
    :2:
    hello2
end:

note: search within line number 33
within: :33:
end:

SLN expressions pitfalls

• When using SLN expressions, there should not be any other symbols before or after exprrssion ( with the exception of using SLN together with between: and within: , see the next paragaphs ).

Following example of correct SLN expressions used with other expressions:

# verify that input text
# has the very first
# hello line
# then any line
# the world line
# then any line 
hello
:1:
world
:3:

This is example of incorrect SLN expressions syntax when SLN expressions are used with other expressions:

# this is wrong
# you cannot use SLN and other
# expressions on the same line
helld
:1: world :3:

• You can use SLN expressions together with between expressions like that ( previous rule of no other expressions on the same line where SLN is written does not apply )

# verify that input text
# between the first
# and tenth line
# has digits
between: { :0: } { :9 }
regexp: \d
end:

# verify that input text
# between the first
# and line with word DONE
# has digits
between: { :0: } { DONE }
regexp: \d
end:

• You can use SLN expressions together with within; expressions like that ( previous rule of no other expressions on the same line where SLN is written does not apply )

# verify that input text
# tenth line
# has digits
within: :9:
regexp: \d
end:

• In task check parser line numbers starts with index 0, so the first line will be :0:, the second will be :1:, etc

Streams

Streams are captures grouped by logical blocks within search context.

While captures and matched lines get flushed with every check step, steams allow to accumulate captures and group them.

Consider an example with captures first:

Input text:

foo
    a
    b
    c
bar

foo
    1
    2
    3
bar

foo
    0
    00
    000
bar

DSL code:

begin:

  regexp: (f) (oo)

  code: <<CODE
  !perl
  print "layer: 1 ", ( join "", map { map {"{$_}"} @{$_}} @{captures()} ), "\n";
  CODE
  
  regexp: (\S+)

  code: print "layer: 2 ", ( join "", map { map {"{$_}"} @{$_}} @{captures()} ), "\n";

  regexp: (\S+)

  code: <<CODE
  !perl
  print "layer: 3 ", ( join "", map { map {"{$_}"} @{$_}} @{captures()} ), "\n";
  CODE
  
  regexp: (\S+)

  code: <<CODE
  !perl
  print "layer: 4 ", ( join "", map { map {"{$_}"} @{$_}} @{captures()} ), "\n";
  CODE
  
  regexp: (bar)

  code: <<CODE
  !perl
  print "layer: 5 ", ( join "", map { map {"{$_}"} @{$_}} @{captures()} ), "\n";
  CODE
  
end:

Output:

[task check] stdout match (s) <(f) (oo)> True
[task check] layer: 1 {f}{oo}{f}{oo}{f}{oo}
[task check] stdout match (s) <(\S+)> True
[task check] layer: 2 {a}{1}{0}
[task check] stdout match (s) <(\S+)> True
[task check] layer: 3 {b}{2}{00}
[task check] stdout match (s) <(\S+)> True
[task check] layer: 4 {c}{3}{000}
[task check] stdout match (s) <(bar)> True
[task check] layer: 5 {bar}{bar}{bar}
    

As we've seen though this example:

• With captures it's impossible to preserve previously matched data and group it by streams ( abc, 123, 0 00 000 ).

• Captures scope is always limited to check step

• Captures are grouped by check steps

• Captures are logical groups ignorant

Streams act differently, they respect search context and logical groups:

• Streams are grouped by search context logical blocks

• Streams allow to accumulate matched data within search context

Consider the same example with streams:

DSL:

begin:

  regexp: (f) (oo)
  regexp: \S+
  regexp: \S+
  regexp: \S+
  regexp: (bar)

  code:  <<CODE
  !perl
    for my $s (@{streams_array()}) {
        my $i=1;
        for my $l (@{$s}){
          print "layer: ", $i++, " ", (join "", map {"{$_}"} @{$l}), "\n";
        }
      print "\n";
    }

  CODE

end:

Output:

[task check] layer: 1 {f}{oo}
[task check] layer: 2 {a}
[task check] layer: 3 {b}
[task check] layer: 4 {c}
[task check] layer: 5 {bar}
[task check] 
[task check] layer: 1 {f}{oo}
[task check] layer: 2 {1}
[task check] layer: 3 {2}
[task check] layer: 4 {3}
[task check] layer: 5 {bar}
[task check] layer: 1 {f}{oo}
[task check] 
[task check] layer: 1 {f}{oo}
[task check] layer: 2 {0}
[task check] layer: 3 {00}
[task check] layer: 4 {000}
[task check] layer: 5 {bar}
    

In this example stream_array() function returns an array of streams.

Every stream is an array of layers, where layer is captured data for a single check step within search context logical block

Streams preserve logical block context. Number of streams relates to the number of successfully matched blocks.

Streams data presentation is much closer to what we see in original text, because it remain original text structure, not breaking it to independent layers.

Stream could be specially useful when combined with range expressions of different range lengths:

Input text:

foo
    2
    4
    6
    8
bar

foo
    1
    3
bar

foo
    0
    0
    0
bar

DSL code:

between: {foo} {bar}

  regexp: (\d+)

  code:  <<CODE
  !perl
    for my $s (@{streams_array()}) {
        my $i=1;
        for my $l (@{$s}){
          print "layer: ", $i++, " ", (join "", map {"{$_}"} @{$l}), "\n";
        }
      print "\n";
    }

  CODE

end:

Output:

[task check] stdout match (r) <(\d+)> True
[task check] layer: 1 {2}
[task check] layer: 2 {4}
[task check] layer: 3 {6}
[task check] layer: 4 {8}
[task check] 
[task check] layer: 1 {1}
[task check] layer: 2 {3}
[task check] 
[task check] layer: 1 {0}
[task check] layer: 2 {0}
[task check] layer: 3 {0}

Streams functions

Following streams functions are implemented for TC compatible languages:

• streams_array

Returns array of streams

• streams()

Hash representation of streams data. Hash keys - unique stream identifiers.

• dump_streams()

Dump streams in human readable format, useful for debugging.

Experimental features

Replace

Replace allows to update file in runtime during check is performed

Following example search HELLO and digits within OK .. DONE range and then increment found digit and update original file:

Input file:

cat << HERE > file.txt
OK
HELLO 1
DONE
HERE

DSL:

between: {OK} {DONE}
    ~regexp: HELLO \s+ (\d+)
end:

code: <<RAKU
!raku
if matched() {
    my $ln = captures-full()[0]<index>;
    my $num = capture()[0].Int;
    $num++;
    replace("file.txt",$ln,"BYE $num");
}
RAKU

Modified file:

OK
BYE 2
DONE

Pay attention that soft check is used here to let main flow continues even though if range search fails

Remove line

remove-line works similarly to replace, but just removing line at number n, following example removes all lines between AAA and BBB, including AAA and BBB lines:

DSL:

between: {AAA} {BBB}
:any:
end:

code: <<RAKU
!raku
for captures-full()<> -> $c {
    remove-line("file.txt",$c<index>)
}
RAKU

Sources

By default STDOUT is a source where input data is read from. One can charge this by using source: directive.

Consider two files - file1.txt and file2.txt with various content. Instead of checking content of both files ( using for example cat file1.txt; cat file2.txt), we can check files independently:

DSL:

source: file1.txt
# all checks are going to be against file1.txt
# data
A1
B1
C1

source: file2.txt
# all checks are going to be against file2.txt
# data
A2
B2
C2

This makes a sense when file1.txt has a following lines:

A1
B1
C1

And file file2.txt respectively:

A2
B2
C2

This is just a very simple example showing the concept, one can do more sophisticated scenarios, for example combining sources with code blocks:

# collect all lines with numbers
# capturing numeric data

~regexp: (\d+)

# save numbers to the file
# numbers.out for
# further processing

code: <<OK
!raku

my @data;

for captures()<> -> $c {
    push @data, $c[0];
}

"numbers.out".IO.spurt(@data.join("\n"));

OK

source: numbers.out

# capture all lines
:any:

# sum all numbers
code: <<OK
!raku

my $sum = 0;

for matched() -> $c {
    $a += Int($c)
}

say "sum: $sum";

OK

This scenario effectively implements linux style pipeline, where the first step extracts all the numbers and the second sums them up.

Sources could be effectively combined with replace and remove-line functions effectively causing reload of data from changed files, consider this example with a failure in the end, proving that the second line has already been deleted:

DSL:

note: === change source ===

source: /tmp/file1.txt
regexp: ^^ A1 $$
B1
C1

note: === change source ===

source: /tmp/file2.txt
regexp: ^^ A2 $$
B2
C2


note: === remove second line from /tmp/file1.txt  ===

code: <<OK
!raku
remove-line("/tmp/file1.txt",1)
OK

note: === change source ===

source: /tmp/file1.txt

B1

Output:

[task stdout]
|> source changed to [/tmp/file1.txt]
21:46:52 :: A1
21:46:52 :: B1
21:46:52 :: C1
|> source changed to [/tmp/file2.txt]
21:46:52 :: A2
21:46:52 :: B2
21:46:52 :: C2
|> source changed to [/tmp/file1.txt]
21:46:54 :: A1
21:46:54 :: C1
[task check]
# === change source ===
stdout match <^^ A1 $$> True
stdout match <B1> True
stdout match <C1> True
# === change source ===
stdout match <^^ A2 $$> True
stdout match <B2> True
stdout match <C2> True
# === remove second line from /tmp/file1.txt  ===
# === change source ===
stdout match <B1> False

FAQ

Q1

Is it common for one line to fail and subsequent lines then fail even though they match?

This only happens if sequential search is used, for example:

task.check

begin:
AAA
B
C
end:

Report:

[task stdout]
23:17:57 :: A
23:17:57 :: B
23:17:57 :: C
[task check]
stdout match (s) <AAA> False
stdout match (s) <B> False
stdout match (s) <C> False
=================
TASK CHECK FAIL

This happens because for block: end: mode, search effectively stops when the first none matching happens, because all the further matches do not matter as we require subsequent match - like AAA, B, C in strict order.

Examples

Following are some selected use cases, for more go to examples/ directory

Apache version

DSL:

regexp: ^^ \s* "Server:" \s+ "Apache/" \d+ "."

Output:

[task stdout]
22:14:15 :: Server: Apache/2.4.62
[task check]
stdout match <^^ \s* "Server:" \s+ "Apache/" \d+ "."> True

Raku Regex Quick Reference

Basic Syntax

Anonymous regex:

/pattern/             # shorthand
rx/pattern/           # explicit
regex { pattern }     # keyword-declared

Named regex:

regex R { pattern }   # named Regex object
token R { pattern }   # implies :ratchet
rule R { pattern }    # implies :ratchet :sigspace

Matching:

"string" ~~ /pattern/     # smartmatch
m/pattern/                # explicit match against $_
"string".match(/pattern/) # method form

Character Classes

Predefined: | Class | Description | |----------------------------------------------------------------------------| | <alpha> | Alphabetic + underscore | | <digit> or \d | Decimal digits | | <xdigit> | Hex digits [0-9A-Fa-f] | | <alnum> | <alpha> + <digit> | | <punct> | Punctuation & symbols | | <space> or \s | Whitespace | | <cntrl> | Control characters | | <lower> / <:Ll> | Lowercase | | <upper> / <:Lu> | Uppercase | | <|wb> | Word boundary | | <ww> | Within word |

Enumerated classes:

<[abc]>        # match a, b, or c
<-[abc]>       # negated: any except a, b, c
<[a..z]>       # range
<[\d] - [13579]>  # set difference

Unicode properties:

<:Script<Latin>>    # Unicode script
<:Block('Basic Latin')>  # Unicode block

Quantifiers

| Quantifier | Meaning | |------------------------------------| | + | One or more | | * | Zero or more | | ? | Zero or one | | ** N | Exactly N times | | ** M..N | Between M and N times | | ** M..* | At least M times |

Captures & Grouping

Positional captures:

/(\d+)/        # captured into $0
/(\d+) (\w+)/  # $0, $1

Named captures:

/$<year>=(\d{4})/  # accessible as $<year>
/<alpha>/           # also captures into named capture

Non-capturing grouping:

[pattern]       # group without capturing

Adverbs & Modifiers

| Adverb | Effect | |-----------------------------------------------| | :i or :ignorecase | Case insensitive | | :r or :ratchet | No backtracking | | :s or :sigspace | Whitespace significant | | :g or :global | Match globally | | :ov or :overlap | Overlapping matches | | :ex or :exhaustive | All possible matches |

Common Patterns

# Date YYYY-MM-DD
\d ** 4 '-' \d ** 2 '-' \d ** 2

# Quoted string
'"' <-[ " ]> * '"'

# Word with optional plural
\w+ s?

# Logical newline
\n

# Any character including newline
\N

# Match start/end of string
^  $   # line boundaries
^^ $$  # logical line boundaries (for multiline)

Integration Guidelines

When using this skill for Sparrow6 task check files:

1. Always verify Raku regex compatibility. Remember that Raku regex differs from Perl/PCRE in significant ways:

   • Spaces are not matched literally
   • No \n in regex patterns within DSL (use sequence expressions for multiline)
   • Use \s+ or " " for whitespace

2. Structure task.check files with proper context management:

   • Start with plain text or regex checks
   • Use begin:/end: for sequences, between:/end: for ranges
   • Place code: or generator: after check expressions to handle captures
   • Always close context modifiers with end:

3. Debugging approach:

   • Use code: expressions to dump matched(), captures() data
   • Verify each check expression independently before combining
   • Remember that failures don't halt execution; check all results

4. Performance considerations:

   • Generator expressions re-execute for each input line; keep them efficient
   • Use streams_array() for accumulating data within context modifiers
   • Prefer plain text checks over regex when exact matching suffices