negate the regexp in validates_format_of

Railsters:

ActiveRecord's validation system puts other database systems to shame.

However, the newbies might not know how to write a regexp that excludes a match, instead of tests for it. Understand - I'm just asking this question to help them. I have been using Regexps since 'grep' on Xenix! But the newbies here might not know how to do this:

  validates_format_of :field_must_break, :with => /\n/

  validates_format_of :field_must_not_break, :with => /^[^\n]$/

The second one fails with data that correctly have no \n linefeed in them.

How do I - I mean "they" - negate a regexp?

Keynan Pratt wrote:

validates_format_of :field_must_not_break, :with => /[^\n]/

That passes on the first non-\n character. /^[^\n]+$/ doesn't work either. (Trying for "all characters between start and end not \n".)

New question: Why don't any variations of Regexp have a bulk negation operator? Most applications can rely on !~.

When we can't use that, or when we must negate a subset of a regexp, the result is super-ugly. The contrapositive of /\n/ should be just as clean-looking, right?

Keynan Pratt wrote:

so you wanted this then? validates_format_of :field_must_not_break, :with => /^[^\n]*$/

assert_match /^[^\n]+$/, "kozmik\nbullfrog"

That assertion passes - we want it to fail.

Next, there are two questions afloat - how to validate (without resorting to validate(), on principle), and whether all regexp libraries everywhere have a big gap.

You can't pass a validating block to validates_exclusion_of; only a proc to determine if the validation should take place. It gets passed the model, so we could put the 'not str.includes?' in there, but that's much more code than simply using a validate().

And 'not str.includes?' doesn't answer the philosophical question about regexps!

Are you looking for a field with only one character [which can’t be a newline]? 'Cause that’s what /[1]$/ does. You might want to add some wildcardier metacharacters to include more possible characters than just one.

Or you might just want to ask this question over and over.

Excuse me, “they” might want to. :wink:

RSL


  1. ^\n ↩︎

Russell Norris wrote:

Are you looking for a field with only one character [which can't be a newline]? 'Cause that's what /^[^\n]$/ does. You might want to add some wildcardier metacharacters to include more possible characters than just one.

Wrong.

Or you might just want to ask this question over and over.

Wrong.

Um… That’s actually right. [About the friggin’ regex].

rsl@sneaky:~$ irb irb(main):001:0> rx = /^a$/ => /^a$/ irb(main):002:0> “test” =~ rx => nil # Doesn’t have an “a”

irb(main):003:0> “tast” =~ rx => nil # Has an “a” but has other characters as well irb(main):004:0> “a” =~ rx => 0 # Surprise! It passes there because it has only “a”

irb(main):005:0> rx = /[1]$/ => /[2]$/ irb(main):006:0> “\n” =~ rx => nil # Fails on returns, as expected irb(main):007:0> “” =~ rx => nil # Fails on empty string

irb(main):008:0> “1” =~ rx => 0 # Passes because there’s only one character irb(main):009:0> “11” =~ rx => nil # Fails because there’s more than one character [

regular expressions - Google Search](regular expressions - Google Search) much?


  1. ^\n ↩︎

  2. ^\n ↩︎

Um… That’s actually right. [About the friggin’ regex].

Sorry, it’s not.

rsl@sneaky:~$ irb irb(main):001:0> rx = /^a$/ => /^a$/ irb(main):002:0> “test” =~ rx => nil # Doesn’t have an “a” irb(main):003:0> “tast” =~ rx => nil # Has an “a” but has other characters as well irb(main):004:0> “a” =~ rx => 0 # Surprise! It passes there because it has only “a” irb(main):005:0> rx = /[1]$/ => /[2]$/ irb(main):006:0> “\n” =~ rx => nil # Fails on returns, as expected irb(main):007:0> “” =~ rx => nil # Fails on empty string irb(main):008:0> “1” =~ rx => 0 # Passes because there’s only one character irb(main):009:0> “11” =~ rx => nil # Fails because there’s more than one character

$ irb

r=/^a$/

=> /^a$/

s=“this\nis\na\nlong\nmulti-line\nstring”

=> “this\nis\na\nlong\nmulti-line\nstring”

puts s

this

is

a

long

multi-line

string

=> nil

s =~ r

=> 8

r2 = /[3]$/

=> /[4]$/

s =~ r2

=> 8

“what?\n!” =~ r2

=> 6

http://www.google.com/search?q=regular+expressions much?

No, 'cause I don’t need to… Do you http://www.ruby-doc.org/docs/ProgrammingRuby/html/language.html#UL ever?

(got your back Philip :wink:

-Rob


  1. ^\n ↩︎

  2. ^\n ↩︎

  3. ^\n ↩︎

  4. ^\n ↩︎

Hi --

Well, snap on my doodle! I was so busy hoping I could give Phlip a heaping bowl of his own brand of snotty reply that I tripped on some of my own brand of over-confidence. Crow meet plate, fork, and knife. [Again.]

I’m still trying to wrap my head around this. I think of “a\nb” as a string containing three characters, a, the newline character, and b. Oh, I think I see… ^ and $ don’t match the end and beginning of a string but the end and beginning of a line… Sigh. Goddangit. Rassin’ frassin’.

Oh well. Apologies to Phlip, I’m fully prepared for and deserving of your next snarky retort. [But not the one after that.]

RSL

Hi --