How to rewrite Regex.Replace (due to async api)

Question

I have a function ReplaceParameters that replaces values in a string by using Regex.Replace. This has been working fine, but now the api that gets replacement-string has become async-only. This is a repro of the current code:

public static string ReplaceParameters(string query)
{
    var result = Regex.Replace(query, @"(?<parameter>\|\w+\|)", ReplaceParameterEvaluator,
                                         RegexOptions.ExplicitCapture);

    return result;
}

private static string ReplaceParameterEvaluator(Match parameterMatch)
{
    var parameter = parameterMatch.Groups["parameter"].Value;
    return GetReplacement(parameter);
}

private static string GetReplacement(string parameter)
{
    //...
}

Since the (new) GetReplacement function now is returning a Task instead of string: private static async Task<string> GetReplacementAsync(string parameter) the ReplaceParameterEvaluator function can't be made compatible with the MatchEvaluator delegate.

Sinc this has to run on a web-server and not cause dead-locks, i cant use any of the dirty async-to-sync hacks like for example this: (use .Result) var replacedQuery = Regex.Replace(query, @"(?<parameter>\|\w+\|)", match => ReplaceParameterEvaluatorAsync(match).Result, RegexOptions.ExplicitCapture);

Is it possible to rewrite the function to find all the texts, and then replace them? Could Regex.Matches be used in some way?

(Seriously though, why is there not a Regex.ReplaceAsync function??)

Either don't use GetReplacementAsync or use "hacks". No elegant solution here. — i3arnon
– i3arnon, Commented Oct 8, 2015 at 11:12
Right after posting i found osdn.jp/projects/opentween/scm/git/open-tween/blobs/master/… - Which contains a simple implementation for RegexAsync.ReplaceAsync. But it does not support capture-groups. Perhaps a simliar aproach could be used in my case? — Steinar Herland
– Steinar Herland, Commented Oct 8, 2015 at 11:18
It seems to first match everything, then find replacements asynchronously and then replace everything. It's using regex twice, but yes, it's probably possible. — i3arnon
– i3arnon, Commented Oct 8, 2015 at 11:27
There is no Regex.ReplaceAsync function because string replacement is not an I/O bound operation, and exposing an async API wouldn't make sense. — Lucas Trzesniewski
– Lucas Trzesniewski, Commented Oct 8, 2015 at 13:23
@LucasTrzesniewski it CAN be an I/O bound operation though, in my case it's replacing <img> tags in HTML with base64 encoded image data read from disk. So - many hanks for your helpful answer below! — Alex from Jitbit
– Alex from Jitbit, Commented Jul 23, 2023 at 18:01

Lucas Trzesniewski · Accepted Answer · 2015-10-08 13:30:54Z

22

It's simple enough to build your own extension method for this:

public static async Task<string> ReplaceAsync(this Regex regex, string input, Func<Match, Task<string>> replacementFn)
{
    var sb = new StringBuilder();
    var lastIndex = 0;

    foreach (Match match in regex.Matches(input))
    {
        sb.Append(input, lastIndex, match.Index - lastIndex)
          .Append(await replacementFn(match).ConfigureAwait(false));

        lastIndex = match.Index + match.Length;
    }

    sb.Append(input, lastIndex, input.Length - lastIndex);
    return sb.ToString();
}

It's straightforward:

Copy unmatched chunks of text as-is to a StringBuilder
Append the result of the callback function where text is matched

This is just a convenience method for calling async callbacks. It's not in the API because regex matching and replacement is a CPU-bound operation, which is naturally not asynchronous.

answered Oct 8, 2015 at 13:30

Lucas Trzesniewski

51.6k11 gold badges115 silver badges169 bronze badges

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

7 Comments

Rawling Over a year ago

You could potentially even parallelize all the calls to replacementFn and then build the string at the end.

Steinar Herland Over a year ago

It's been some years since I last used the regex-classes... As you see in the original code, I'm using Regex.Replace (The static version) - Is it possible to change my code to use a Regex instance instead?

Lucas Trzesniewski Over a year ago

new Regex(@"(?<parameter>\|\w+\|)", RegexOptions.ExplicitCapture).ReplaceAsync(...)

svick Over a year ago

@Rawling I would be careful with that. What if replacementFn is not thread-safe?

Steinar Herland Over a year ago

Tanks! I have no problem with the regex not being cpu-bound. My function however is. - And you gave a nice workaround.

|

Collectives™ on Stack Overflow

How to rewrite Regex.Replace (due to async api)

1 Answer 1

7 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Related