2

I am writing a JSON parser for .NET and it parses JSON objects near perfectly so far. One problem I'm having, though, is that it will parse simple strings, but will not parse complex strings. Here is an example:

It will parse \"Hi there!\" as a string.

It will not parse \"Hi !*\t\r\n,,{}]][] (.&^.)@!+=~`' there\"

The spec I am using for a JSON string is directly from the JSON website.

My .NET regex strings (as I interpreted from the site) are:

string json_char = @"(\\""|\\\\|\\/|\\b|\\f|\\n|\\r|\\u|[^(\""|\\)])";
string json_string = @"(\""" + json_char + @"*\"")";

The above are exactly as they appear in Visual Studio. Note that with the @ symbols, two double-quotes ("") are required to specify a single double-quote (") character in the actual string value.

The above regex strings match nothing in the second, complex string example I gave above. I've fiddled with the regex strings, but nothing seems to work.

What I want is a regex string that will parse a JSON string as specified by the website. Any help is appreciated.

Improve this question
7
  • 4
    not a good idea, instead use a JSON library like Json.NET james.newtonking.com/pages/json-net.aspx Commented Dec 22, 2010 at 22:43
  • In addition, suggestions as to why the above regex strings do not work would be helpful. Commented Dec 22, 2010 at 22:44
  • Why don't you use process of elimination to figure out exactly which characters are breaking the regex and then add them into json_char? Commented Dec 22, 2010 at 22:45
  • @Mauricio Scheffer, my first though too, but Mike might just be doing it for fun or learning. Commented Dec 22, 2010 at 22:46
  • 2
    @tster - Bingo. It is primarily a learning experience. Is there an easier way to match " a control character or any UNICODE character except " and \ ..."? That is what I was going for in my 'json_char' regex string. Commented Dec 22, 2010 at 22:49

3 Answers 3

Reset to default
2

If I were writing the parser, I might approach it a bit differently. Parsing is a different kind of operation than matching, and sometimes a Regex may only get you half-way there. For example, I would probably match and capture all the name/value pairs from the parent JSON document using a Regex like this: string pattern = @"(?:""[^""\\]*(?:\\.[^""\\]*)*"")+"; which will return everything between and including the opening and closing quotes of the string. Then I would check the captured string for the exceptional cases outlined in the JSON spec, like a backslash not being followed by a valid escape code, and then throw an exception if I found any issues. I might also consider replacing any naked escape codes, like a tab character, with a \t. Once I had the captured string sanitized, and error-checked, I could run Regex.Unescape() to return the final string.

Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

Agreed. I haven't got extensive experience with making parsers, but I had a very positive experience with a compiler-compiler called Coco/R. AFAIU this one is unique because of what's called attrbuted grammar files and semantic actions. The first basically allows you to define input and output parameters to methods generated to parse some part of a language construct, and the second let's you insert custom C# code wherever you want in the generated parser, which is handy when you regenerate. It can even step into grammar files when debugging! scifac.ru.ac.za/coco/cshcoco.htm
2

The first thing you should do is get rid of all those unneeded backslashes. Some of them should just be removed; for example, the backslash in \"" is just ignored. The rest of the backslashes are pulling their weight, but you don't have to write them out every time. For example, this will match escaped quotes and backslashes plus the whitespace escape sequences (FYI, you left the t out of your regex):

@"\\[""\\/bfnrt]"

I left out the u for Unicode escapes because it has to be followed by four hexadecimal digits; you have to match them separately from the other escapes. Adding them to the above regex gives you

@"\\(?:[""\\/bfnrt]|u[0-9A-Fa-f]{4})"

Finally, you seem to be using [^(""|\\)] for the catch-all part, i.e., any Unicode character except a quotation mark, a backslash, or a control character. What that part actually matches is any Unicode character except (, ", |, ), or a backslash. The correct way to match anything but a quotation mark or a backslash would be [^""\\], but you also need to exclude control characters. For that you can use Unicode property, \p{Cc}. Here's the whole thing:

@"""(?:[^\p{Cc}""\\]+|\\(?:[""\\/bfnrt]|u[0-9A-Fa-f]{4}))*"""

Notice that I included the quote delimiters in this regex instead of adding them in a separate step as you did. I'm assuming the backslash in \" is not meant to be treated as a literal character; otherwise you would have used two of them.

Improve this answer

Comments

0

Note that with the @ symbols, two double-quotes ("") are required to specify a single double-quote (") character in the actual string value.

Additionally to this, in an @-ed string, a backslash character is a literal backslash. So if you write, say, @"\\t", the regex engine will look for a backslash followed by a "t", not a tab character.

I suspect that these superfluous backslashes are the source of your problem.

Improve this answer

1 Comment

I believe he really is trying to match the actual escape sequence, not a tab character.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.