I have a little experience working with .hl7 data (format for transmitting health data), so I decided to try and write an .hl7 parser using fparsec.

A standard .hl7 segment is a single line with:

1. First 3 chars are the header
2. Next 5 chars are the separators (typically |^~&)
3. Fields / repetitions delimited by pipes

A segment has fields or repetitions (repeating fields). A field can have components. A component can have sub-components.

Sample .hl7 message header:

MSH|^~\\&|Location|SendingApp|DateTimeOfMessage 

I am looking for feedback on:

1. Data design - the types
2. The functions - am I doing something really dumb? How can I make this cleaner?
3. How to not end up with empty fields (i.e. if a field doesn't have any data, I don't want it in the list on the segment).

type Subcomponent = {value:string; position:int} type Component = {subcomponents: Subcomponent list; position:int} type SingleField = {components: Component list; position:int} type Field = Repetitions of SingleField list | SingleField of SingleField type Segment = { name:string; fields:Field list; } type Hl7Message = { segments:Segment list } let hl7Seps = "|^~\\&" let normalChar = noneOf hl7Seps let unescape c = match c with | 'F' -> '|' | 'R' -> '~' | 'S' -> '^' | 'T' -> '&' | 'E' -> '\\' | c -> c let escapedChar = attempt (pchar '\\' >>. anyChar |>> unescape .>> skipChar '\\') <|> pchar '\\' let pHl7Element = manyChars (normalChar <|> escapedChar) let pcomp = sepBy pHl7Element (pchar '&') |>> (fun vals -> List.mapi (fun i s -> {value = s; position = i}) vals) let pfield = sepBy pcomp (pchar '^') |>> (fun comps -> List.mapi (fun i c -> {subcomponents = c; position = i}) comps) let pRepsOrField = sepBy pfield (pchar '~') |>> (fun fields -> match fields.Length with | 0 | 1 -> SingleField {components = fields.Item 0; position = 0} | _ -> Repetitions (List.mapi (fun i c -> {components = c; position = i}) fields)) let pheader = anyString 3 |>> (fun name -> name) let pSegment = pipe2 pheader (sepBy pRepsOrField (pchar '|')) (fun name repsOrFields -> {name = name; fields = repsOrFields}) test pSegment "EVN|A&1^B^C|123~456~789"