jaynetics / js_regex Goto Github PK

View Code? Open in Web Editor NEW

69.0 69.0 11.0 489 KB

Converts Ruby regexes to JavaScript regexes.

License: MIT License

Ruby 100.00%

javascript regular-expression ruby

js_regex's People

Contributors

Stargazers

Watchers

Forkers

mrcthms owst mojavelinux rroger singpolyma tagliala gogainda monsido gust royzwambag igor-drozdov

js_regex's Issues

possible to change python regex to js regex

Allow regexp_parser 2.0

There are 4 failures about multiple quantifiers:

Failures:

  1) JsRegex::Converter::Base#convert when there are multiple quantifiers drops adjacent/multiplicative fixes ({x}) without warning
     Failure/Error:
       expect(/a{4}{6}/).to\
       become(/a{6}/)
     
       expected a{6}, got (?:a{4}){6}
     # ./spec/lib/js_regex/converter/base_spec.rb:165:in `block (4 levels) in <top (required)>'

  2) JsRegex::Converter::Base#convert when there are multiple quantifiers drops adjacent/multiplicative ranges ({x,y}) without warning
     Failure/Error:
       expect(/a{2,4}{3,6}/).to\
       become(/a{3,6}/)
     
       expected a{3,6}, got (?:a{2,4}){3,6}
     # ./spec/lib/js_regex/converter/base_spec.rb:170:in `block (4 levels) in <top (required)>'

  3) JsRegex::Converter::Base#convert when there are multiple quantifiers drops mixed adjacent quantifiers without warning
     Failure/Error:
       expect(/ab{2,3}*/).to\
       become(/ab*/)
     
       expected ab*, got a(?:b{2,3})*
     # ./spec/lib/js_regex/converter/base_spec.rb:175:in `block (4 levels) in <top (required)>'

  4) JsRegex::Converter::Base#convert when there are multiple quantifiers drops multiple adjacent quantifiers without warning
     Failure/Error:
       expect(/ab{2}{3}{4}{5}/).to\
       become(/ab{5}/)
     
       expected ab{5}, got a(?:(?:(?:b{2}){3}){4}){5}
     # ./spec/lib/js_regex/converter/base_spec.rb:180:in `block (4 levels) in <top (required)>'

Finished in 0.81114 seconds (files took 0.2056 seconds to load)
297 examples, 4 failures

Failed examples:

rspec ./spec/lib/js_regex/converter/base_spec.rb:164 # JsRegex::Converter::Base#convert when there are multiple quantifiers drops adjacent/multiplicative fixes ({x}) without warning
rspec ./spec/lib/js_regex/converter/base_spec.rb:169 # JsRegex::Converter::Base#convert when there are multiple quantifiers drops adjacent/multiplicative ranges ({x,y}) without warning
rspec ./spec/lib/js_regex/converter/base_spec.rb:174 # JsRegex::Converter::Base#convert when there are multiple quantifiers drops mixed adjacent quantifiers without warning
rspec ./spec/lib/js_regex/converter/base_spec.rb:179 # JsRegex::Converter::Base#convert when there are multiple quantifiers drops multiple adjacent quantifiers without warning

May refer to: e3da3f3

Wrap errors from regexp_parser

Thanks for an excellent gem! 💎 Works splendidly!

One thing I've noticed a few times is that errors from the underlying gem regexp_parser can bubble up.

Would it make sense for js_regex to wrap them in its own error classes?

Something like:

begin
  # …
rescue Regexp::Parser::ParserError, Regexp::Syntax::SyntaxError, … => err
  raise JsRegex::ParsingError, "Underlying parser library regexp_parser failed, #{err.message} (#{err.class})"
end

At some level I'd argue that errors raised by a library is also part of the public API, and right now the API of one of the dependencies is "bleeding through". Just a small thing though, not a major problem in practice.

For a simpler solution, could just update the readme and list the known base error classes that regexp_parser can raise.

Regression / Breaking change in v3.1.0?

Hi, I'm using your gem as a dependency of https://github.com/DavyJonesLocker/client_side_validations

After upgrading from 3.0.0 to 3.1.0, a test stopped working:

https://github.com/DavyJonesLocker/client_side_validations/blob/v11.1.3/test/core_ext/cases/test_core_ext.rb#L40-L43

Failure:
CoreExtTest#test_regexp_remove_group_options [/Users/geremia/dev/client_side_validations/test/core_ext/cases/test_core_ext.rb:42]:
--- expected
+++ actual
@@ -1 +1 @@
-{:source=>"(something)", :options=>"g"}
+{:source=>"(?:something)", :options=>"g"}

Could you please tell me if I need to fix the test case or if there is a regression?

The commit which breaks tests is a35129f

JSRegex ignores dollar signs

JsRegex.new(/\$/).to_h

outputs:

{:source=>"", :options=>"g"}

Quantifier {,m} incorrectly translated

hi! I think I've found a bug but I'm no expert in regexes so I'll just describe what I'm experiencing

in Ruby this regex is valid and works

Regexp.new("[a-z]{,3}").match("hey")[0]
=> "hey"

when we translate it to JS using our js_regex we get

::JsRegex.new(Regexp.new("[a-z]{,3}")).to_h.slice(:source, :options)
=> {:source=>"[a-z]{,3}", :options=>""}

but when I try it in Chrome, it matches nothing

"hey".match(new RegExp("[a-z]{,3}"))
null

however if I change the quantifier to {1,3} instead of just {,3}

"hey".match(new RegExp("[a-z]{1,3}"))
['hey', index: 0, input: 'hey', groups: undefined]

it works. but if we now use the u flag

var regexp = new RegExp("[a-z]{,3}", "u")
Uncaught SyntaxError: Invalid regular expression: /[a-z]{,3}/: Incomplete quantifier
    at new RegExp (<anonymous>)
    at <anonymous>:1:14

it fails! let's replace again {,3} with {1,3}

"hey".match(new RegExp("[a-z]{1,3}", "u"))
['hey', index: 0, input: 'hey', groups: undefined]

and it works.

according to this doc a quantifier cannot be written as {,m}
so maybe js_regex should translate {,m} as {1,m}?

ES6 mode

The readme mentions:

Possible future improvements might include an "ES6 mode" using the u flag, which would allow for more concise representations of astral plane properties and sets.

I was wondering if this improvement is still planned, or if the gem is considered complete at this point.

Remove the javascript 'g' regex-option as default for all ruby regexes?

Hey, first of all: nice gem, I like using it!

I struggled a bit with the 'g' option. if it is set in javascript.

var re = /pattern/g;
re.test('has_pattern_in_it'); # true
re.test('has_pattern_in_it'); # false
re.test('has_pattern_in_it'); # true
re.test('has_pattern_in_it'); # false

See documentation
this is not really intuitive, so my question is: Is there a reason why the g option is set as a default or would it be okay to remove it?

Escaping issue for character classes containing a /

> JsRegex.new(/[\/]/).to_s
=> "/[/]/"

expected: "/[\/]/" or "/[\x2F]/"

Is lookbehind support possible?

Hi there

Would it be possible to add (optional?) lookbehind support?

I'm unclear whether it's technically hard to do, or it's due to lack of lookbehind support in JS land. As of Sept 2022, Can I Use is showing Safari as the only mainstream browser not supporting lookbehind, with support in Chrome, Edge, Firefox and Opera: https://caniuse.com/js-regexp-lookbehind. FYI I found the relevant webkit bug: https://bugs.webkit.org/show_bug.cgi?id=174931

Saw a reference in #10, but thought I'd raise as a new issue for visibility.

Many thanks for a such a fantastic library. I've been using it with no issues for a while, so was a little sad when I saw my first js_regex warning of "Dropped unsupported lookbehind assertion".

Thanks

Greg

3.10.0 removes escape from hyphen in character range causing unexpected matches

Recently updated from 3.9.0 to 3.10.0.

Issue noticed when trying to parse: '\-\. (match the individual symbols apostrophe ', hyphen - and point .)
When parsed by js_regex it became '-. which now matches all the symbols between apostrophe ' and point ..

Previous versions parsed '\-\. to '\x2D. which works as intended.

MRE results:

js_regex_version: 3.9.0
regex string = '\-\.
ruby regexp = (?-mix:\A['\-\.])
js_regex = /^['\x2D.]/
warnings: []

vs.

js_regex_version: 3.10.0
regex string = '\-\.
ruby regexp = (?-mix:\A['\-\.])
js_regex = /^['-.]/
warnings: []

Validation mode

It's not possible for oniguruma and this module to work together in a performant manner atm.
We'd be doing both regexes each time we shell out to oniguruma instead and it'd happen for every use of \G.

Instead we could try to add validation for this.
Basically turn everything except warn_of_unsupported_feature into a no-op and check if the warning stack is empty after..

Output incorrect when matching on backslashes and using Regexp::union

js_regex 2.2.0, Ruby 2.3.1.

[1] pry(main)> JsRegex.new(/\//).to_s
=> "/\\//g" # Correct
[2] pry(main)> JsRegex.new(Regexp.union(/\//)).to_s
=> "/\\//g" # Correct
[3] pry(main)> JsRegex.new(Regexp.union(/\//, /a/)).to_s
=> "/(\\\\/)|(a)/g" # Incorrect, should be "/(\\/)|(a)/g".
[4] pry(main)> Regexp.union(/\//, /a/)
=> /(?-mix:\/)|(?-mix:a)/
[5] pry(main)> JsRegex.new(/(?-mix:\/)|(?-mix:a)/).to_s
=> "/(\\/)|(a)/g" # Oddly, back to being correct.