File types:
- v2: DOS alternate B
- v3: ProDOS INT with aux type in range [$1000,$3fff]
- v4/v5: ProDOS INT with aux type in range [$4000,$5fff]
Primary references:
- v2: reverse engineering, primarily by Andy McFadden
- v3: LISA v3.1 source code, from A2ROMulan CD-ROM
- v4/v5: Lisa816 v5.0a (433) source code, from A2ROMulan CD-ROM
LISA is an assembler for Apple II code written by Randall Hyde. The name is pronounced "LE ZA" rather than "LI SA".
The program kept the source file in tokenized form, which reduced the memory footprint and the time required to assemble code. This is true for all versions, although the file format evolved significantly over time.
Version 3.x was the last for 8-bit computers. Versions 4 and 5 were branded "LISA 8/16", and added support for the Apple IIgs.
The file structure is fairly straightforward:
+$00 / 2: program version? - usually $1800, which might indicate v2.4 (2.4 * 2560)
+$02 / 2: length of file data; needed because DOS 3.3 doesn't store this for alternate B
+$04 /nn: series of lines
Each line is:
+$00 / 1: line length (does not include length byte)
+$01 /nn: data
+$xx / 1: end-of-line marker ($0d)
The last line in the file has a length of 255 ($ff). The last length byte is followed by two more $ff bytes.
The line data is a mix of plain text and tokens. The tokenization was primarily applied to the instruction mnemonics.
Lines starting with '*' or ';' are full-line comments. Lines starting with an upper-case letter are code lines that start with a label, lines starting with a caret (^) have a local (numeric) label, and lines starting with a space do not have a label. All other characters are invalid at the start of a line. If present, regular labels always take 8 bytes in the source file, and local labels take 7 bytes, padded with trailing spaces if necessary. If the label is terminated with a colon, it will be included at the end.
If the label is on a line by itself, the next character will be the end-of-line marker. If not, the next byte will be the encoded opcode or pseudo-op mnemonic, with a value >= 0x80.
The opcode is followed by a byte that indicates the contents of the operand field. It will be $00
if there is no operand, $01 for absolute addressing, $02 for immediate, $05 for direct page
indirect indexed Y), and so on. Operands that should be taken literally, e.g.
ASC "HELLO, WORLD!", use $20. The text of the operand follows.
If the line has a comment, it will be preceded by $bb (high-ASCII semicolon). Because that value cannot appear in a valid operand, the end of the operand can be identified unambiguously, and thus it's not necessary to track open/close quotes when converting lines to text. Lines without comments will follow the operand with $0d.
Auxtypes $17fc, $1800, $1ffc, $2000, $26fc, and $27fc have been seen.
The file format looks like this:
+$00 / 2: length of code section, in bytes
+$02 / 2: length of symbol table, in bytes
+$04 /nn: symbol table, 0-512 entries, 8 bytes each
+$xx /yy: code
Symbol table entries are 8 bytes each:
+$00 / 6: 8-character label, packed into 6 bytes
+$06 / 2: value? (usually zero)
The code section is a series of lines with a complex encoding. All symbols are encoded as references to the symbol table, so the text of each symbol is only stored once. Numeric values and operators used in operands are encoded.
At least one file has been found that appears to use a different table of opcode mnemonics. The file "anix.equates" appears in some ANIX distributions, and has the same filetype and auxtype as other source files. However, even though it appears to decode successfully, all of the opcodes are incorrect. Curiously, the ANIX 2.1 command "LPRINT" can successfully convert "anix.equates" to text, but generates incorrect output for LISA v3.1 source files.
Auxtypes $40e8, $40e9, and $50e1 have been seen. A filename suffix of ".A" is commonly used.
The file format looks like this:
+$00 / 2: version number
+$02 / 2: offset of first byte past end of symbol table
+$04 / 2: symbol count
+$06 / 1: tab position for opcodes
+$07 / 1: tab position for operands
+$08 / 1: tab position for comments
+$09 / 1: CPU type
+$0a / 6: (reserved)
+$10 /nn: symbol table
+$xx /yy: code
The symbol table is a series of strings that have a preceding length and are null terminated. The length byte includes the null termination and the length byte itself.
The code section is a series of lines with a complex encoding, similar in overall structure to v3 but extended.