Kaitai Struct use an field (u1) as type chooser or be part of the next attribute (str)

[Kaipheu]

Hello, thanks for that project very useful !

I've tried for 1 day to describe a part of a file format.

Two u1 "del" are used as delimiter and to know the structure of the data between. Those blocks are chained, like this:

struct : del|content|del|del|content|del...

values : 01|AC4A55|01|03|0022AA|03...

So I can describe that as:

seq:
  - id: delimiter_start
    type: u1
    enum: content_enum_type
  - id: content
    type: 
      switch-on: delimiter_start
      cases:
      "content_enum_type::one":   type_one
      "content_enum_type::two":   type_two
      "content_enum_type::three": type_three
  - id: delimiter_stop
    type: u1
types:
  type_one:
    seq:
     -id: A
      type: u1
     -id: B
      type: u1
     -id: C
      type: u1
  type_two:
    seq:
     -id: D
      type: u2
     -id: E
      type: u1

  type_three:
    seq:
     -id: F
      type: u4
enum:
   content_enum_type:
      0x1: type_one
      0x2: type_two
      0x3: type_three

This is working well, but I have one case where we have to store UTF-8 strings because they haven't delimiters, i.e.: del|content|del|string_content|del|content|del

So it's here I'm blocked, I need to read the delimiter_start but eventually not storing it.

I've tried to use instances but I can't have the position of delimiter_start, I can't make a substream to deal with relative pos because the line of content isn't fixed (can't use size) and the terminator could be two different values (can't use terminator).

Is there a solution I'm missing or maybe a new feature has to be implemented?
I have tried several things but am always stuck in the fact if delimiter_start is read I can't "suppress" resulting in :
delimiter_start = H string = ELLO WORLD.

I also tried this

...
  - id: delimiter_start
    type: u1
    enum: content_enum_type
    if delimiter_start == content_enum_type::one
#    if delimiter_start == content_enum_type::two
#    if delimiter_start == content_enum_type::three
...

But this doesn't work, the compiled JS looks like:

...
if(this.delimiterStart == myType.content_enum_type.ONE){
   this.delimiterStart = this._io.readU1();
}
...

this.delimiterStart is undefined before the readU1 so the if is always false.

[GreyCat]

Sounds like your "stopping delimiter" is present in some cases and missing in some others. Why not include it into types (e.g. type_one, type_two, type_three), where it is necessary, but omit it from the UTF-8-related type where it's not needed?

[Kaipheu]

Hello, thank you to answer !
Yes a can do it but the field which haven't delimiters will have its first byte parse as delimiters and not as content of the field.
Some thing like : delimiter_start = H, string = ELLO WORLD.