The grammar is described in src/net/haspamelodica/javazmach/assembler/parser/grammar.txt. The important basics should be covered here.

The following instructions are weirdly encoded and not currently supported:

inc, dec, inc_chk, dec_chk, store, pull, load, jump

.ZHEADER <field> <value>

The following header fields are set automatically and do not need to be explicitely specified:

FILE_LENGTH, FILE_CHECKSUM, VERSION, ALPHABET_TABLE, HIGH_MEM_BASE, DICTIONARY, OBJ_TABLE, GLOBAL_VAR_TABLE, STATIC_MEM_BASE

For this to work, the object table, globals and dictionary must be declared somewhere. They can, however, be left empty. The AlphabetTable is not yet supported and automatically set to 0.

Labels and variable names must be unique. In particular, routine locals do not shadow globals, their names must differ. Even worse, locals in different routines may not have the same name. This is due to routines having no defined end point, hence no clear scoping of variables is possible. An exception is made for names declared in macros.

<opcode mnemonic> [<argument1>[, <argument2>[, ...]]] [? [~] <branch target>] [-> <store target>]

Note that the condition of a conditional jump can be negated by prepending a tilde to the branch target, i.e., the instruction jz var ?~ label jumps to label if var is not zero.

Macros can be called similarly to routines, but instead of performing a routine call, they simply insert their content wherever they were invoked (as you would expect from a macro). Named labels/locations/variables declared inside a macro are limited to that macros scope. Hence, when declaring a routine in a macro, the locals do not clash with locals declared elsewhere. If a macro A invokes another macro B, then labels/names declared in A can also be used in B.

.MACRO
.MACROPARAM <param1 name>
.MACROPARAM <param2 name>
.MACROPARAM <param3 name>
...

// content

.ENDMACRO

Parameters that reference a value are expanded with ..<param name>.

Parameters that define a new name (e.g., label declarations) are expanded with .:<param name>

Macros are invoked with .<macro name> [<param1>[, <param2>[, ...]]]

.VALUE <ident> <expr>

Values are named constants. They can be arithmetic expressions of constants and other values. Values can be used in most places, like instruction operands, buffer size, attribute and property indices. Note: they cannot currently be used as the value of a property declaration. There is no good reason for this, we just have not gotten around to it yet.

.BUFFER [<N> BYTE] <value> [<value>]

A buffer is basically just a named region of memory. The size or value must be specified, otherwise the size cannot be deduced.

.DICTIONARY [<separator1> [<separator2> [...]]]
.KEY <key1> [.DATA <size1> BYTE <VALUE1> [.DATA ...]]
.KEY <key2> [.DATA <size2> BYTE <VALUE2> [.DATA ...]]
...

Token separator characters are specified as char literals, e.g., '+'. You do not need to specify any, if the default separators are sufficient for you. Keys can be specified as string literals.

.OBJECTS
.PROPERTY <index> <byte sequence>
.PROPERTY <index> <byte sequence>
...

.OBJECT <ident> "<obj1_name>" {
    .ATTRIBUTE <index>
    ...

    .PROPERTY <index> <byte sequence>
    ...

    .OBJECT <ident> "<obj1_child1_name>" {
        ...
    }
    ...
}

Properties declared outside of objects are default property values. If a default property is not explicitly specified, the default value is 0. Objects have an identifier, which can be used to refer to them as operands to instructions. In addition, the name is how they are emitted by the print_obj istruction. Objects declared in other objects are their children in the object tree.

.GLOBALS
.GLOBAL <global1> [<default1>]
.GLOBAL <global2> [<default2>]
.GLOBAL <global3> [<default3>]
...

.ROUTINE <routine name>
.LOCAL <local1> [<default1>]
.LOCAL <local2> [<default2>]
.LOCAL <local3> [<default3>]
...

// body

Not yet supported.

Not yet supported. Header field should be set to zero.