Searching text with reserved characters with fuzzy

[DaveG (David Groves)]

Hi All,

I have queries where I want to allow searching on part numbers/help ticket numbers and the like with reserved characters in them. For example:
SR-1234
SR/1234

The fields are analyzed with the standard analyzer and with an edge ngrams N=2 to 20.

When I query like this:

    {
        "query": {
            "bool": {
                "should": [
                    {
                        "simple_query_string": {
                            "query": "sr",
                            "fields": [
                                "PartNo", "PartNo.autocomplete"
                            ]
                        }
                    }
                ]
            }
        },
        "highlight": {
            "fields": {
                "PartNo": {}
            }
        }
    }

I get a result, the highlighter shows matches such as:
sr -> "SR/1234"
sr-1234 -> "SR/1234"

and this makes sense. The analyzer at index and search time has removed the "-", "/" etc and indexed sr and 1234 independently, and they are found independently.

Now, if I search with fuzzy by adding the ~:
sr~ -> "SR/1234". ok,
but
sr-1234~ -> no results
sr\-1234~ -> no results

Can someone explain what fuzzy is doing here and if it is possible to formulate a query to achieve some level of fuzziness over the number (which is likely to be a ticket number/bug number and quite large, 8 digits or more)?
I know the edge-ngrams will help with trailing uncertainty in the number (missing trailing characters), but if a user transposes 2 characters it will not help.

all thoughts appreciated,
Dave.