[Ruby, PHP] Errors with slashes in pattern

[scop]

Description

For this in YAML syntax:

pattern: '^a/b$'

...the PHP code generated is:

if (!preg_match("/^a/b$/", ...

...and for Ruby:

Regexp.new(/^a/b$/)

The problem with both is that the slash between a and b should be backslash escaped to avoid errors. I tried this in DefaultCodegen.addRegularExpressionDelimiter:

return "/" + pattern.replace("/", "\\/") + "/";

...but it makes no difference on the eventual output.

Swagger-codegen version

Current master

[wing328]

Your fix should work.

scala> var a = "^a/b$"
a: String = ^a/b$

scala> a.replace("/", "\\/")
res0: String = ^a\/b$

Please submit a PR and I'll do some tests

[scop]

I agree it should work, but as said, in my tests it did not end up resulting any changes to those problematic ruby and php code spots -- for some reason, the slash remained unescaped there. Anyway, #3707 is the PR.

[wing328]

@scop I did some tests with your PR and have no idea either as it failed to escape forward slash.

@cbornet when you've time, I wonder if you can take a look to see if you've any clue why the replace does not work as intended

[ePaul]

I debugged this a bit ... the problem might be that the result of this method gets later passed (by toRegularExpression) through DefaultCodegen.escapeText, which does (between others) .replace("\\/", "/")) – i.e. it undoes the transformation.

I traced the transformation of the regexp through all the methods called from DefaultCodegen.toRegularExpression (in a PhpClientCodegen instance). The result is the same in master as in #3707.

method	output (#3707)	output (master)
(input)	[A-Z]/.*	^[A-Z]/.*
addRegularExpressionDelimiter	/[A-Z]\/.*/	/^[A-Z]/.*/
escapeJava(...)	/[A-Z]\\/.*/	/^[A-Z]/.*/
.replace("\\/", "/"))	/^[A-Z]\/.*/	/^[A-Z]/.*/
unescapeJava(...)	/^[A-Z]/.*/	/^[A-Z]/.*/
.replaceAll("[\\t\\n\\r]", " ")	/^[A-Z]/.*/	/^[A-Z]/.*/
.replace("\\", "\\\\")	/^[A-Z]/.*/	/^[A-Z]/.*/
.replace("\"", "\\\"")	/^[A-Z]/.*/	/^[A-Z]/.*/
escapeUnsafeCharacters(...)	/^[A-Z]/.*_/	/^[A-Z]/.*_/
.trim()	/^[A-Z]/.*_/	/^[A-Z]/.*_/

I'm not sure what the idea of this escapeText method in DefaultCodegen is – it unescapes, replaces, escapes again, replaces some more, and again escapes. It looks quite brittle, and is likely not made for regular expressions (as we can see by the last _ inserted, which breaks the whole expression.)

[wing328]

@ePaul thanks for the detailed troubleshooting results. We'll try to address this after the 2.2.2 release.