Parsing EDIFACT with Clojure's Instaparse

Want to discuss the content of this article? Join the conversation on Twitter!

The structure of EDIFACT files

EDIFACT is a standardized text-based format that enables companies to send business messages between each other (EDI).


Here’s a small example, to give you an idea of what EDIFACT look like.

NAD+BY+++Buyer company+Street name+City'
LIN+1++Article #4232:SA'

This entire thing is called a message. On the second line, you can read the message type: ORDERS, a purchase order message.

Each line is called a segment. The line-breaks are optional, but convenient for displaying and reading.

A segment consists of

It is possible to define your own delimiters, using an UNA segment in the beginning of the file. We will use the default delimiters here.

Let’s break one of the segments into it’s pieces.

LIN+1++Article #4232:SA'

The tag of this segment is LIN. It has three elements.

  1. The number “1”.
  2. An empty element.
  3. A composite element of two parts:
    1. “Article #4243”
    2. “SA”

To interpret what the meaning of the different elements are, we look in the EDIFACT specification which tells us that this segment is a line item and the item number is 1. The third element is the item number identification. “SA” is the item number code which stands for supplier’s article number, which in our case was “Article #4243”.

Parsing in Clojure: Instaparse

In the Clojure world, there exists this very nice parser library called Instaparse. It aims to let you describe your grammar using standard EBNF notation, and turn it into an executable parser.

Step by step, let us turn our informal description of EDIFACT segments into simple EBNF notation.

EBNF grammar

A message consist of one or more segments.

message ::= segment+

A segment is a tag, followed by one or more elements, ending with a terminator and optionally a new line.

segment ::= tag element+ terminator newline?

Finally, an element is just a element delimiter followed by an element component. In turn, a component is either simple, or composite, and a composite component is two or more simple values separated by the component delimiter.

element        ::= elem-delim elem-component
elem-component ::= simple | composite
composite      ::= simple (component-delim simple)*
simple         ::= chars

There we have it. In fact, this is very close to how the parser is described. The only addition I have made is to let chars be either escaped chars or ordinary chars to account for escaping (using the symbol ? (question mark)). E.g. if a text field contains a colon, it should be preceded by a question mark (?:).

Clojure version of the EBNF grammar

Let’s have a look at the definition for our parser in Clojure.

;; You need to require instaparse.core as insta, see Quickstart in
;; the Instaparse documentation (
(def parser (insta/parser
             "message = segment+
              segment = tag element+ terminator newline?
              tag = #'[A-Z0-9]{3}'
              element = elem-delim elem-component
              <elem-component> = simple | composite

              <composite> = simple (component-delim simple)*
              <simple> = chars

              chars = (escaped-char | char)*
              escaped-char = #'\\?.'
              <char> = #'[^+:\\'\\?]+'

              <elem-delim> = <'+'>
              <component-delim> = <':'>
              <newline> = <'\\r'?> <'\\n'>
              <terminator> = <\"'\">"))

The angle brackets on the left side is used for hiding tags in the resulting tree, and angle brackets on the right hand side is for hiding values. For example, the delimiters are hidden entirely.

In order to preserve the element components correctly in the presence of escaped characters, the tags for chars and escaped-char are left in the parse tree.

The can be collapsed neatly using insta/transform.

(defn run-parser
  (->> s
        {:chars str
         :escaped-char second})))

Let’s have a look at a couple of examples.

(run-parser "PAI+::42'")
;; =>
[:message [:segment [:tag "PAI"] [:element "" "" "42"]]]

(run-parser "FTX+REG+++Information?::about:regulatory:authority:'")
;; =>
[:message [:segment [:tag "FTX"] [:element "REG"] [:element ""] [:element ""] [:element "Information:" "about" "regulatory" "authority" ""]]]

NAD+BY+++Buyer company+Street name+City'
LIN+1++Article #4232:SA'
;; =>
 [:segment [:tag "UNB"] [:element "UNOC" "3"]
  [:element "SE1212121212" "ZZZ"] [:element "DE3434343434" "ZZZ"]
  [:element "150728" "0000"] [:element "1234567"]]
 [:segment [:tag "UNH"] [:element "1"] [:element "ORDERS" "D" "96A" "UN"]]
 [:segment [:tag "BGM"] [:element "220"] [:element "100"]]
 [:segment [:tag "DTM"] [:element "4" "20150701" "102"]]
 [:segment [:tag "NAD"] [:element "BY"] [:element ""] [:element ""] [:element "Buyer company"] [:element "Street name"] [:element "City"]]
 [:segment [:tag "LIN"] [:element "1"] [:element ""] [:element "Article #4232" "SA"]]
 [:segment [:tag "QTY"] [:element "47" "40"]]
 [:segment [:tag "UNS"] [:element "S"]]
 [:segment [:tag "CNT"] [:element "2" "1"]]
 [:segment [:tag "UNT"] [:element "9"] [:element "1"]]
 [:segment [:tag "UNZ"] [:element "1"] [:element "1234567"]]]

If you have any comments or feedback, don’t hesitate to send me an email or a message on Twitter @lexicallyscoped.

←  Go Back