NAME
Outthentic::DSL
SYNOPSIS
Outthentic::DSL - language to verify (un)structured text.
Install
$ cpanm Outthentic::DSL
Developing
$ git clone https://travis-ci.org/melezhik/outthentic-dsl
$ cd outthentic-dsl
$ perl Makefile.PL && make && make install
Glossary
Input text
An arbitrary, often unstructured text being verified. It could be any text.
Examples:
html code
xml code
json
plain text
emails :-)
http headers
another program languages code
Outthentic DSL
Is a language to verify arbitrary text
Outthentic DSL is both imperative and declarative language
Declarative way
You define rules ( check expressions ) to describe expected content.
Imperative way
You extend a process of verification using regular programming languages - like Perl, Bash and Ruby, see examples below.
DSL code
A program code written on Outthentic DSL language to verify text input.
Search context
Verification process is taken in a context.
By default search context is equal to an original text input stream.
However a search context might be changed in some situations ( see within, text blocks and ranges expressions ).
DSL parser
DSL parser is the program which:
parses DSL code
parses text input
verifies text input ( line by line ) against a check expressions ( line by line )
Verification process
Verification process consists of matching lines of text input against check expressions.
This is schematic description of the process:
For every check expression in a check expressions list:
* Mark this check step in `unknown' state.
* For every line in input text:
* Verify if it matches check expression. If line matches then mark step in `succeeded' state.
* Next line.
End of lines loop.
* If the check step marked in `unknown' state, then mark it in `failed' state.
* Next check expression.
End of expressions loop.
Check if all check steps are succeeded. If so then input text is considered verified, else - not verified.
A final presentation of verification results should be implemented in a certain client using parser api and not being defined at this scope.
Parser API
Outthentic::DSL provides program API for client applications.
This is example of verification some text against 2 lines;
use Outthentic::DSL;
my $otx = Outthentic::DSL->new(<<HERE, { debug_mod => 0 });
Hello
My name is Outthentic!
HERE
$otx->validate(<<'CHECK');
Hello
regexp: My\s+name\s+is\s+\S+
CHECK
print "status\tcheck\n";
print "==========================\n";
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
status check
==========================
true text has 'Hello'
true text match /My\s+name\s+is\s+\S+/
Methods list:
new
This is constructor to create an Outthentic::DSL instance.
Obligatory parameters are:
text
input text to get verified
Outthentic::DSL->new("Hi! Welcome to my birthday party.\nLet's have a fun" );
Optional parameters are passed as hash:
match_l - truncate check expressions to a
match_l
bytes when generating results
This is useful when debugging long check expressions:
use Outthentic::DSL;
my $otx = Outthentic::DSL->new( 'A'x99 , { match_l => 9 });
$otx->validate('A'x99);
print "status\tcheck\n";
print "==========================\n";
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
status check
==========================
true text has 'AAAAAAAAA'
Default value is 40
.
debug_mod - enable debug mode
Possible values is one of:
0,1,2,3,4
Set to 1 or 2 or 3 or 4 if you want to see some debug information appeared at console.
Increasing debug_mod value results in more low level information appeared.
Default value is
0
- means do not emit debug messages.
validate
Perform verification process.
Obligatory parameter is:
a string with DSL code
Example:
$otx->validate(<<'CHECK');
# there should be digits
regexp: \d
# and greetings
regexp: hello \s+ \w+
CHECK
results
Returns validation results as array containing { type, status, message } hashes.
Outthentic clients
Client is a external program using DSL API. Existed Outthentic clients:
Swat - web application testing tool
Outthentic - multipurpose scenarios framework
DSL code syntax
Outthentic DSL code comprises following entities:
Comments
Blank lines
Check expressions:
plain strings
regular expressions
text blocks
within expressions
asserts expressions
validator expressions
range expressions
Code expressions
Generator expressions
Check expressions
Check expressions define patterns to match against an input text stream.
Here is a simple example:
use Outthentic::DSL;
my $otx = Outthentic::DSL->new(<<'HERE');
HELLO
HELLO WORLD
My birth day is: 1977-04-16
HERE
$otx->validate(<<'CHECK');
HELLO
regexp: \d\d\d\d-\d\d-\d\d
CHECK
print "status\tcheck\n";
print "==========================\n";
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
status check
==========================
true text has 'HELLO'
true text match /\d\d\d\d-\d\d-\d\d/
There are two basic types of check expressions:
Plain text expressions
Plain text expressions define a lines an input text to contain.
use Outthentic::DSL;
my $otx = Outthentic::DSL->new(<<'HERE');
I am ok, really
HELLO Outthentic !!!
HERE
$otx->validate(<<'CHECK');
I am ok
HELLO Outthentic
CHECK
print "status\tcheck\n";
print "==========================\n";
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
status check
==========================
true text has 'I am ok'
true text has 'HELLO Outthentic'
Plain text expressions are case sensitive:
use Outthentic::DSL;
my $otx = Outthentic::DSL->new(<<'HERE');
I am ok
HERE
$otx->validate(<<'CHECK');
I am OK
CHECK
print "status\tcheck\n";
print "==========================\n";
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
status check
==========================
false text has 'I am OK'
Regular expressions
Similarly to plain text matching, you may require that input lines match some regular expressions.
This should be Perl Regular Expressions.
Example:
use Outthentic::DSL;
my $otx = Outthentic::DSL->new(<<'HERE');
2001-01-02
Name: Outthentic
App Version Number: 1.1.10
HERE
$otx->validate(<<'CHECK');
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
CHECK
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
true text match /\d\d\d\d-\d\d-\d\d/
true text match /Name:\s+\w+/
true text match /App Version Number:\s+\d+\.\d+\.\d+/
One or many?
Parser does not care about how many times check expression matches an input text.
If at least one line in a text matches the check expression - this check is considered as successful.
If you use capturing regex expressions, parser accumulates all captured data to make it possible further processing.
Example:
use Outthentic::DSL;
use Data::Dumper;
my $otx = Outthentic::DSL->new(<<'HERE');
1 - for one
2 - for two
3 - for three
HERE
$otx->validate(<<'CHECK');
regexp: (\d+)\s+-\s+for\s+(\w+)
CHECK
print Dumper($otx->{captures});
Output:
[
[
'1',
'one'
],
[
'2',
'two'
],
[
'3',
'three'
]
]
See "captures" section for full explanation of a captures mechanism.
Comments, blank lines and text blocks
Comments and blank lines don't impact verification process but you may use them for the sake of DSL code readability.
Comments
Comment lines start with #
symbol, comments are ignored by parser.
DSL code:
# comments could be represented at a distinct line, like here
The beginning of story
Hello World # or could be added for the existed expression to the right, like here
Blank lines
Blank lines are ignored as well.
DSL code:
# every story has the beginning
The beginning of a story
# then 2 blank lines
# end has the end
The end of a story
But you can't ignore blank lines in a text blocks, see text blocks subsection for details.
Use :blank_line
marker to match blank lines inside text blocks.
DSL code:
# :blank_line marker matches blank lines
# this is especially useful
# when match in text blocks context:
begin:
this line followed by 2 blank lines
:blank_line
:blank_line
end:
Text blocks
Sometimes you need to match a text against a sequence of lines like in code below.
use Outthentic::DSL;
my $otx = Outthentic::DSL->new(<<'HERE');
this string followed by
that string followed by
another one string
with that string
at the very end.
HERE
$otx->validate(<<'CHECK');
# this text block
# consists of 5 strings
# going consecutive
begin:
# plain strings
this string followed by
that string followed by
another one
# regexp patterns:
regexp: with\s+(this|that)
# and the last one in a block
at the very end
end:
CHECK
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
true [b] text has 'this string followed by'
true [b] text has 'that string followed by'
true [b] text has 'another one'
true [b] text match /with\s+(this|that)/
true [b] text has 'at the very end'
A negative example:
my $otx = Outthentic::DSL->new(<<'HERE');
that string followed by
this string followed by
another one string
with that string
at the very end.
HERE
$otx->validate(<<'CHECK');
# this text block
# consists of 5 strings
# going consecutive
begin:
# plain strings
this string followed by
that string followed by
another one
# regex patterns:
regexp: with\s+(this|that)
# and the last one in a block
at the very end
end:
CHECK
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
true [b] text has 'this string followed by'
false [b] text has 'that string followed by'
true [b] text has 'another one'
true [b] text match /with\s+(this|that)/
true [b] text has 'at the very end'
begin:
, end:
markers decorate text blocks content.
Markers should not be followed by any text at the same line.
Don't forget to close the block ...
Be aware if you leave "dangling" begin:
marker without closing end:
parser will remain in a text block mode till the end of the file, which is probably not you want:
DSL code:
begin:
here we begin
and till the very end
of this text
we remain in `text block` mode
Code expressions
Code expressions are just a pieces of 'some language code' you may inline and execute during parsing process.
By default, if language is no set Perl language is assumed. Here is example:
use Outthentic::DSL;
my $otx = Outthentic::DSL->new('hello');
$otx->validate(<<'CHECK');
hello
code: print "hi there!\n";
CHECK
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
hi there!
true text has 'hello'
As you may notice code expression here has no impact on verification process, this trivial example just shows that you may inline some programming languages code into Outthentic DSL. See generators section on how dynamically create new check expressions using common programming languages.
You may use other languages in code expressions, not only Perl.
Use here
document style ( see multiline expressions section ) and proper shebang to insert code written in other languages. Here are some examples:
perl5
code: <<HERE
!perl
print 'hi there!'
HERE
bash
code: <<HERE
!bash
echo 'hi there!'
HERE
ruby
code: <<CODE
!ruby
puts 'hi there!'
CODE
Asserts
Asserts expressions consists of assert value, and description - a short string to describe assert.
Assert value should be something to be treated as false or true in Perl, here is examples:
DSL code
# you may have assert expressions as is
# then assert value should be Perl value to be treated as true or false
#
assert: 0 this is not true in Perl
assert: 1 this is true in Perl
assert: "OK" none empty string is for true in Perl
assert: "" empty string is for false in Perl
Asserts almost always to be created dynamically with generators. See the next section.
Generators
Generators is the way to generate new Outthentic entities on the fly.
Generator expressions like code expressions are just a piece of code to be executed.
The only requirement for generator code - it should return new Outthentic entities.
If you use Perl in generator expressions ( which is by default ) - last statement in your code should be one of three:
reference to array of strings where every string should represent a new Outthentic entity
string to represent a new Outthentic entities, this could be multiline string
path to file with check expressions - new outthentic entities
Usually first two options most convenient way.
If you use languages other than Perl to produce new Outthentic entities you should print them into stdout. See examples below.
A new Outthentic entities are passed back to parser and executed immediately.
Generators expressions start with generator:
marker.
Here is simple example.
use Outthentic::DSL;
my $otx = Outthentic::DSL->new('HELLO');
$otx->validate(<<'CHECK');
generator: [ 'H', 'E', 'L', 'O' ];
CHECK
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
Output:
true text has 'H'
true text has 'E'
true text has 'L'
true text has 'O'
If you use other languages to generate expressions, you just need to print entries into stdout. Here are some generators examples for other languages:
Original check expressions list:
Say
HELLO
This generator creates 3 new check expressions:
generator: <<CODE
!bash
echo say
echo hello
echo again
CODE
Or if you prefer Ruby:
generator: <<CODE
!ruby
puts 'say'
puts 'hello'
puts 'again'
CODE
Updated check list:
Say
HELLO
say
hello
again
Here is more complicated example using Perl language.
# this generator creates
# comments
# and plain string check expressions:
use Outthentic::DSL;
my $otx = Outthentic::DSL->new(<<'HERE');
foo value
bar value
HERE
$otx->validate(<<'CHECK');
generator: <<CODE
my %d = ( 'foo' => 'foo value', 'bar' => 'bar value' );
join "\n", map { ( "# $_" , $d{$_} ) } keys %d;
CODE
CHECK
for my $r (@{$otx->results}) {
print $r->{status} ? 'true' : 'false', "\t", $r->{message}, "\n";
}
A check list being generated:
# foo
foo value
# bar
bar value
Output:
true text has 'foo value'
true text has 'bar value'
Generators could produce not only check expressions but code expressions and ... another generators.
So ... use your imagination power! ...
This is fictional example.
Input Text:
A
AA
AAA
AAAA
AAAAA
DSL code:
generator: <<CODE
sub next_number {
my $i = shift;
$i++;
return if $i>=5;
[
'regexp: ^'.('A' x $i).'$',
"generator: next_number(".$i.")"
]
}
CODE
Generators are commonly used to create an asserts.
This is short example for Ruby language:
number: (\d+)
generator: <<CODE
!ruby
puts "assert: #{capture()[0] == 10}, you've got 10!"
CODE
Validators
WARNING!!! You should prefer asserts over validators. Validators feature will be deprecated soon!
Validator expressions are perl code expressions used for dynamic verification.
Validator expressions start with validator:
marker.
A Perl code inside validator block should return array reference.
Once code is executed a returned array structure treated as:
first element - is a status number ( Perl true or false )
second element - is a helpful message
Validators a kind of check expressions with check logic expressed in program code. Here is examples:
# this is always true
validator: [ 10>1 , 'ten is bigger then one' ]
# and this is not
validator: [ 1>10, 'one is bigger then ten' ]
# this one depends on previous check
regexp: credit card number: (\d+)
validator: [ captures()[0][0] == '0101010101', 'I know your secrets!' ]
# and this could be any
validator: [ int(rand(2)) > 1, 'I am lucky!' ]
Validators are often used in conjunction with the captures expressions. This is another example.
Input text:
# my family ages list
alex 38
julia 32
jan 2
DSL code:
# let's capture name and age chunks
regexp: /(\w+)\s+(\d+)/
validator: <<CODE
my $total=0;
for my $c (@{captures()}) {
$total+=$c->[0];
}
[ ( $total == 72 ), "total age" ]
CODE
Multiline expressions
Multilines in check expressions
When parser parses check expressions it does it in a single line mode :
check expression is always single line string
input text is parsed in line by line mode, thus every line is validated against a single line check expression
Here is example.
Input text:
Multiline
string
here
DSL code:
# check list
# always
# consists of
# single line expressions
Multiline
string
here
regexp: Multiline \n string \n here
Results:
+--------+---------------------------------------+
| status | message |
+--------+---------------------------------------+
| OK | matches "Multiline" |
| OK | matches "string" |
| OK | matches "here" |
| FAIL | matches /Multiline \n string \n here/ |
+--------+---------------------------------------+
Use text blocks if you want to represent multiline checks.
Multilines in code expressions, generators and validators
Perl expressions, validators and generators could contain multilines expressions
There are two ways to write multiline expressions:
using back slash delimiters to split multiline string to many chunks
using HERE documents expressions
Back slash delimiters
\
delimiters breaks a single line text on a multi lines.
Example:
# What about to validate stdout
# With sqlite database entries?
generator: \
use DBI; \
my $dbh = DBI->connect("dbi:SQLite:dbname=t/data/test.db","",""); \
my $sth = $dbh->prepare("SELECT name from users"); \
$sth->execute(); \
my $results = $sth->fetchall_arrayref; \
[ map { $_->[0] } @${results} ] \
HERE documents expressions
Is alternative to make your multiline code more readable:
# What about to validate stdout
# With sqlite database entries?
generator: <<CODE
use DBI;
my $dbh = DBI->connect("dbi:SQLite:dbname=t/data/test.db","","");
my $sth = $dbh->prepare("SELECT name from users");
$sth->execute();
my $results = $sth->fetchall_arrayref;
[ map { $_->[0] } @${results} ]
CODE
Captures
Captures are pieces of data get captured when parser validate lines against a regular expressions:
Input text:
# my family ages list.
alex 38
julia 32
jan 2
# let's capture name and age chunks
regexp: /(\w+)\s+(\d+)/
code: << CODE
for my $c (@{captures}){
print "name:", $c->[0], "\n";
print "age:", $c->[1], "\n";
}
CODE
Data accessible via captures():
[
['alex', 38 ]
['julia', 32 ]
['jan', 2 ]
]
Usually captured data is good candidates for assert checks.
DSL code:
generator: << CODE
!ruby
total=0
captures().each do |c|
total+=c[0]
end
puts "assert: #{total == 72} 'total age of my family'"
CODE
captures() function
captures() function returns an array reference holding all the chunks captured during latest regular expression check.
Here is another example:
# check if stdout contains lines
# with date formatted as date: YYYY-MM-DD
# and then check if first date found is yesterday
regexp: date: (\d\d\d\d)-(\d\d)-(\d\d)
generator: <<CODE
use DateTime;
my $c = captures()->[0];
my $dt = DateTime->new( year => $c->[0], month => $c->[1], day => $c->[2] );
my $yesterday = DateTime->now->subtract( days => 1 );
my $true_or_false = (DateTime->compare($dt, $yesterday) == 0);
[
"assert: $true_or_false first day found is - $dt and this is a yesterday"
];
CODE
capture() function
capture() function returns a first element of captures array.
it is useful when you need data related only first successfully matched line.
DSL code:
# check if text contains numbers
# a first number should be greater then ten
regexp: (\d+)
generator: [ "assert: ".( capture()->[0] > 10 )." first number is greater than 10 " ]
Search context modificators
Search context modificators are special check expressions which not only validate text but modify search context.
By default search context is equal to original input text stream.
That means parser executes validation use all the lines when performing checks
However there are two search context modificators to change this behavior:
within expressions
range expressions
Within expressions
Within expression acts like regular expression - checks text against given patterns
Text input:
These are my colors
color: red
color: green
color: blue
color: brown
color: back
That is it!
DSL code:
# I need one of 3 colors:
within: color: (red|green|blue)
Then if checks given by within statement succeed next checks will be executed in a context of succeeded lines:
# but I really need a green one
green
The code above does follows:
try to validate input text against regular expression "color: (red|green|blue)"
if validation is successful new search context is set to all matching lines
These are:
color: red
color: green
color: blue
thus next plain string checks expression will be executed against new search context
Results:
+--------+------------------------------------------------+
| status | message |
+--------+------------------------------------------------+
| OK | matches /color: (red|green|blue)/ |
| OK | /color: (red|green|blue)/ matches green |
+--------+------------------------------------------------+
Here more examples:
# try to find a date string in following format
within: date: \d\d\d\d-\d\d-\d\d
# we only need a dates in 2000 year
2000-
Within expressions could be sequential, which effectively means using &&
logical operators for within expressions:
# try to find a date string in following format
within: date: \d\d\d\d-\d\d-\d\d
# and try to find year of 2000 in a date string
within: 2000-\d\d-\d\d
# and try to find month 04 in a date string
within: \d\d\d\d-04-\d\d
Speaking in human language chained within expressions acts like specifications.
When you may start with some generic assumptions and then make your requirements more specific. A failure on any step of chain results in immediate break.
Range expressions
Range expressions also act like search context modificators - they change search area to one included between lines matching right and left regular expression of between statement.
It is very similar to what Perl range operator does when extracting pieces of lines inside stream:
while (<STDOUT>){
if /foo/ ... /bar/
}
Outthentic analogy for this is range expression:
between: foo bar
Between statement takes 2 arguments - left and right regular expression to setup search area boundaries.
A search context will be all the lines included between line matching left expression and line matching right expression.
A matching (boundary) lines are not included in range.
These are few examples:
Parsing html output
Input text:
<table cols=10 rows=10>
<tr>
<td>one</td>
</tr>
<tr>
<td>two</td>
</tr>
<tr>
<td>the</td>
</tr>
</table>
DSL code:
# between expression:
between: <table.*> <\/table>
regexp: <td>(\S+)<\/td>
# or even so
between: <tr.*> <\/tr>
regexp: <td>(\S+)<\/td>
Multiple range expressions
Multiple range expressions could not be nested, every new between statement discards old search context and setup new one:
Input text:
foo
1
2
3
FOO
100
BAR
bar
FOO
10
20
30
BAR
DSL code:
between: foo bar
code: print "# foo/bar start"
# here will be everything
# between foo and bar lines
regexp: \d+
code: <<CODE
for my $i (@{captures()}) {
print "# ", $i->[0], "\n"
}
print "# foo/bar end"
CODE
between: FOO BAR
code: print "# FOO/BAR start"
# here will be everything
# between FOO and BAR lines
# NOT necessarily inside foo bar block
regexp: \d+
code: <<CODE
for my $i (@{captures()}) {
print "#", $i->[0], "\n";
}
print "# FOO/BAR end"
CODE
Output:
# foo/bar start
# 1
# 2
# 3
# 100
# foo/bar end
# FOO/BAR start
# 100
# 10
# 20
# 30
# FOO/BAR end
Restoring search context
And finally to restore search context use reset\_context:
statement.
Input text:
hello
foo
hello
hello
bar
DSL code:
between foo bar
# all check expressions here
# will be applied to the chunks
# between /foo/ ... /bar/
hello # should match 2 times
# if you want to get back to an original search context
# just say reset_context:
reset_context:
hello # should match three times
Range expressions caveats
Range expressions can't verify continuous lists.
That means range expression only verifies that there are some set of lines inside some range. It is not necessary should be continuous.
Example.
Input text:
foo
1
a
2
b
3
c
bar
DSL code:
between: foo bar
1
code: print capture()->[0], "\n"
2
code: print capture()->[0], "\n"
3
code: print capture()->[0], "\n"
Output:
1
2
3
If you need check continuous sequences checks use text blocks.
Experimental features
Below is highly experimental features purely tested. You may use it on your own risk! ;)
Streams
Streams are alternative for captures. Consider following example.
Input text:
foo
a
b
c
bar
foo
1
2
3
bar
foo
0
00
000
bar
DSL code:
begin:
foo
regexp: (\S+)
code: print '#', ( join ' ', map {$_->[0]} @{captures()} ), "\n"
regexp: (\S+)
code: print '#', ( join ' ', map {$_->[0]} @{captures()} ), "\n"
regexp: (\S+)
code: print '#', ( join ' ', map {$_->[0]} @{captures()} ), "\n"
bar
end:
Output:
# a 1 0
# b 2 00
# c 3 000
Notice something interesting? Output direction has been inverted.
The reason for this is Outthentic check expression works in "line by line scanning" mode when text input gets verified line by line against given check expression.
Once all lines are matched they get dropped into one heap without preserving original "group context".
What if we would like to print all matching lines grouped by text blocks they belong to?
As it's more convenient way ...
This is where streams feature comes to rescue.
Streams - are all the data successfully matched for given group context.
Streams are applicable for text blocks and range expressions.
Let's rewrite last example.
DSL code:
begin:
foo
regexp: \S+
regexp: \S+
regexp: \S+
bar
code: <<CODE
for my $s (@{stream()}) {
print "# ";
for my $i (@{$s}){
print $i;
}
print "\n";
}
CODE
end:
Stream function returns an arrays of streams. Every stream holds all the matched lines for given logical block.
Streams preserve group context. Number of streams relates to the number of successfully matched groups.
Streams data presentation is much closer to what was originally given in text input:
Output:
# foo a b c bar
# foo 1 2 3 bar
# foo 0 00 000 bar
Stream could be specially useful when combined with range expressions of various ranges lengths.
For example.
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
for my $s (@{stream()}) {
print "# ";
for my $i (@{$s}){
print $i;
}
print "\n";
}
CODE
Output:
# 2 4 6 8
# 1 3
# 0 0 0
Examples
Some code examples mostly mentioned at this documentation could be found at
examples/
directoryA plenty of other examples could be found at Outthentic module
Environment variables
I'll document these variables later. Here is just a list:
OUTKEEPSOURCE_FILES
OTX_DEBUG
Author
Home page
https://github.com/melezhik/Outthentic-dsl
COPYRIGHT
Copyright 2016 Alexey Melezhik.
This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
See also
Alternative Outthentic DSL introduction could be found here - intro.md
Thanks
To God as the One Who inspires me to do my job!