Be sure you start with the right documentation, which must be purchased from ASC X12. Documentation from a trading partner might be usable in a pinch, but is often missing details or has details relevant only to that partner. If you are unsure, the cover page for the document should look like this:

Tables

The overall structure of the grammar is given in section 2.2 of the documentation (be sure it is marked IMPLEMENTATION and not STANDARD). The top-level tables can be written as TableDef.header, TableDef.detail, and TableDef.summary.

Note for some reason, some grammars show the SE segment at the end of a detail table, rather than in its own summary table. The SE segment is the end of the document, so when you code up the grammar, you should create a TableDef.summary to contain the SE segment (see 837P). This is because a detail table can be repeated, while a summary table cannot, and the SE segment should only occur once in a document.

Loops

Loops are groups of segments which have a particular starting segment that indicates the start of the group, followed by other segments. Loops may be nested, and each loop may be repeated a specified number of times. You can translate each loop in the documentation to a LoopDef.build(name, repeat_count, *children) where children are either segments or child loops.

In the example above, the "2000C PATIENT HIERARCHY" loop has a repeat count of RepeatCount.unbounded, child segments HL and PAT, and many child loops beginning with "2000CA PATIENT NAME". Since HL is the first segment in the loop, when the parser reads an HL in the right context, it will then create the corresponding loop in the parse tree.

Segments

Segments are sort of like a struct in the C programming language, where the fields are called elements. Unlike structs, each time a segment occurs in the grammar, it can be given a different name, different flags to indicate which fields are required or optional, different set of allowed values in each field, and a different number of allowed repeats.

One of the most commonly used segments is NM1, which is generally some kind of name. In the above example, you an NM1 Patient Name, but there are many others in the grammar for different people.

The important parts of the image above have probably already been transcribed in an existing SegmentDef (unless you find a segment that hasn't been defined), which are the same across all grammars (all the X12 attributes). The "loop", "usage", and "repeat count" were described earlier, but are again shown here.

Segments can be described using BuilderDsl.Segment(offset, id, name, usage, repeat_count, *elements). The first argument comes from the Pos. # column in the table diagram above. For instance, the NM1 Patient Name segment has an offset of 150. The usage argument is denoted in a few places: on the above diagrams, it is the labeled Usage; this indicates if the segment is required or optional, or if its presence depends on other conditions.

Elements

Elements are (apart from composite elements), some kind of atomic data type, like a string, number, date, etc. Each place a segment is specified in the grammar, its element properties are also specified. Beware that each place that a segment appears can have different details about its elements, per-occurence.

This diagram indicates some of the properties of each element, in this particular instance of the NM1 segemnt. The bold border around an element indicates that it is Required, while strike-through text indicates a NotUsed. If you encounter a segment that hasn't already been defined, some of the other properties needed to define the segment are also included on the diagram, but we'll ignore those for now.

The third section of each element detail gives a name to each element, usually in the gray section labeled IMPLEMENTATION NAME, and also specifies its usage, allowed values, and minimum and maximum lengths. The usage indicator is in the first column, name and allowed values in the fourth. Minimum and maximum length, and number of decimal places are given in the last column, but these rarely need to be specified as they are usually the same as the default declared in the ElementDef.