Formats

Memory Layout

Layout encoding

Binary serialization

Except for a small header, the binary serialization of a tree is identical to its representation in memory.

Header format

packet-beta
    title Binary Header (Start)
    0-7: "$"
    8-15: "X"
    16-23: "M"
    24-31: "L"
    32-39: "Format Major"
    40-47: "Format Minor"
    48-55: "Configs"
    56-57: "Endian"
    58-63: "reserved-0"
    64-69: "size (bits) delta_ptr_t"
    70-75: "size (bits) xml_size_t"
    76-81: "size (bits) xml_count_t"
    82-87: "size (bits) xml_enum_size_t"
    88-95: "reserved-1"
    96-111: "documents count"
    112-127: "reserved-2"
    128-191: "symbols size"

An array of sections follows. The definition of a section depends on the data types of the build:

struct __attribute__ ((packed)) section_t{
    struct __attribute__ ((packed)){
        delta_ptr_t  base;      //Aligned to the beginning of the symbols table.
        xml_count_t  length;    //Aligned to base
    } name;
    delta_ptr_t     base;      //Aligned to start_data() (after symbols)
    xml_count_t     length;    //Relative to base
}

packet-beta
    title Configs
    0-2: "Symbols"
    3-4: "raw_strings"
    5-6: "allow_comments"
    7-8: "allow_procs"

Multi-document archives

Multi-document archives are based on the binary format introduced before.
The document count will just not be 1, multiple sections are going to be present, whose names are stored in the shared table of symbols.

Indices

TBW.