Optimize Regex

[TonyValenti]

Is there an API that can optimize/minify the string representation of a regex?
For example, if I had:
(aaaaaa1)|(aaaaaa2)|(aaaaaa3)
it would be nice to get something like:
(aaaaaa)(1|2|3)

Tagging @stephentoub because he seems to be doing a ton of work on regex source gen stuff.

[stephentoub]

For example, if I had:
(aaaaaa1)|(aaaaaa2)|(aaaaaa3)
it would be nice to get something like:
(aaaaaa)(1|2|3)

That's not a valid transformation, as it changes the number of capture groups and what each capture group would capture.

Now, if your question was instead about aaaaaa1|aaaaaa2|aaaaaa3, then .NET 7 will optimize that to the equivalent of aaaaaa[123], which you can see by looking at the code generated by the source generator:

                // Description:
                // ○ Match the string "aaaaaa".
                // ○ Match a character in the set [1-3].
    
                protected override bool FindFirstChar()
                {
                    int pos = base.runtextpos, end = base.runtextend;
                    global::System.ReadOnlySpan<char> inputSpan = base.runtext;
                    
                    if (pos < end - 6)
                    {
                        int i = global::System.MemoryExtensions.IndexOf(inputSpan.Slice(pos, end - pos), "aaaaaa");
                        if (i >= 0)
                        {
                            base.runtextpos = pos + i;
                            return true;
                        }
                    }
                    
                    // No starting position found
                    NoStartingPositionFound:
                    base.runtextpos = end;
                    return false;
                }
    
                [global::System.Runtime.CompilerServices.SkipLocalsInit]
                protected override void Go()
                {
                    global::System.ReadOnlySpan<char> inputSpan = base.runtext;
                    int pos = base.runtextpos, end = base.runtextend;
                    int original_pos = pos;
                    global::System.ReadOnlySpan<byte> byteSpan;
                    global::System.ReadOnlySpan<char> slice = inputSpan.Slice(pos, end - pos);
                    
                    if ((uint)slice.Length < 7)
                    {
                        goto NoMatch;
                    }
                    
                    // Match the string "aaaaaa".
                    {
                        byteSpan = global::System.Runtime.InteropServices.MemoryMarshal.AsBytes(slice);
                        if (global::System.Buffers.Binary.BinaryPrimitives.ReadUInt64LittleEndian(byteSpan) != 0x61006100610061ul ||
                            global::System.Buffers.Binary.BinaryPrimitives.ReadUInt32LittleEndian(byteSpan.Slice(8)) != 0x610061u)
                        {
                            goto NoMatch;
                        }
                    }
                    
                    if ((((uint)slice[6]) - '1' > (uint)('3' - '1'))) // Match a character in the set [1-3].
                    {
                        goto NoMatch;
                    }
                    
                    // The input matched.
                    pos += 7;
                    base.runtextpos = pos;
                    base.Capture(0, original_pos, pos);
                    return;
                    
                    // The input didn't match.
                    NoMatch:;
                }

but there's no API nor plans for an API to round-trip that back to a valid string representation.

[TonyValenti]

Thanks @stephentoub !