Bad Escape Error parsing regex

[corylanou]

> select * FROM /^payments\./
ERR: error parsing query: bad escape:  at line 1, char 26

[dgnorton]

david [5:58 PM]
Regex syntax question for everyone. Currently, /foo\./ returns a Bad Escape error. I can fix that by either forcing the user to enter /foo\\./ or by disallowing / to ever appear in a regex. Thoughts? Alternative ideas? (edited)

otoolep [6:00 PM]
@david: I go with escaping. You never know what someone might want.

david [6:02 PM]
@otoolep: I should also say that all escapes have to be double escaped. e.g., if you wanted to find the literal text cpu.* you would have to enter /cpu\\.\\*/

david [6:04 PM]
@hottoddy: how common do you think it will be for users to need to match / in a regex?

david [6:05 PM]
e.g., do we have many users using measurement names or tag values like foo/bar?

david [6:10 PM]
If we disallow / in a regex, they could still use \x2F to match that char. http://play.golang.org/p/0OUeTVBsvC

/cc @benbjohnson

[beckettsean]

\x2F for the edge case feels like an easier thing to maintain and document than double-escaping everything always. I vote we make the 80% case easy (especially since it's really more the 99% case.)

[pauldix]

Why can't we escape / just like with other escapes? e.g. /foo\/bar/, /foo\\bar/, /foo\.bar/

Double escaping sounds very bad to me. I think it'll be a usability nightmare.

[dgnorton]

Because it's going through two parsers...InfluxQL and then Go's regex parser. Both use \ for escapes. In order to pass \ through to the regex parser, it currently has to be double escaped.

Since InfluxQL regexp uses / as the delimiter, we could disallow that char in regexes and then not check for escapes during InfluxQL parsing of regexes. The down side to this is that users would have to use \x2F to match / but I don't know how common that is.

[pauldix]

For the InfluxQL parser, couldn't it just look at the next character after a / to see if it's a \? If it is, skip over that character and continue parsing. The InfluxQL parser should leave the escapes in so they can be passed on to the Go regex parser.