The Hobbit: Routine at 6F47

Used by the routine at TOKENISE.

Copies the word from the input line, turning each letter into its 5-bit code with AND $1F -- which works because 'A' is $41 and the codes were chosen to be the low five bits of the ASCII -- and stops at the first character below $40, so punctuation and spaces end a word without being tested for.

Then the index: the first letter doubled and added to WORD_INDEX gives the bucket's offset, and that added to WORD_INDEX again gives the first entry. From there each call unpacks one candidate -- the tokeniser comes back in at TRY_ENTRY for the next -- and the bucket is over when an entry's initial letter stops matching the one typed, which is its only end marker.

Used by the routine at TOKENISE. Copies the word from the input line, turning each letter into its 5-bit code with AND $1F -- which works because 'A' is $41 and the codes were chosen to be the low five bits of the ASCII -- and stops at the first character below $40, so punctuation and spaces end a word without being tested for. Then the index: the first letter doubled and added to WORD_INDEX gives the bucket's offset, and that added to WORD_INDEX again gives the first entry. From there each call unpacks one candidate -- the tokeniser comes back in at TRY_ENTRY for the next -- and the bucket is over when an entry's initial letter stops matching the one typed, which is its only end marker.
MATCH_WORD	6F47	LD DE,TYPED_LETTERS	Copy the typed word as 5-bit codes to TYPED_LETTERS, up to the first character below $40
	6F4A	LD B,$00
MATCH_WORD_0	6F4C	LD A,(HL)
	6F4D	CP $40
	6F4F	JR C,MATCH_WORD_1
	6F51	AND $1F
	6F53	LD (DE),A
	6F54	INC DE
	6F55	INC HL
	6F56	INC B
	6F57	JR MATCH_WORD_0
MATCH_WORD_1	6F59	LD A,B	Keep its length
	6F5A	LD (TYPED_LENGTH),A	Keep its length
	6F5D	PUSH HL	IX = the first word in the dictionary under its initial letter
	6F5E	LD HL,(TYPED_LETTERS)
	6F61	LD H,$00
	6F63	LD DE,WORD_INDEX
	6F66	ADD HL,HL
	6F67	ADD HL,DE
	6F68	LD E,(HL)
	6F69	INC HL
	6F6A	LD D,(HL)
	6F6B	LD IX,WORD_INDEX
	6F6F	ADD IX,DE
	6F71	POP HL
This entry point is used by the routine at TOKENISE.
TRY_ENTRY	6F72	LD (DICTIONARY_ENTRY),IX	Remember which entry is being tried
This entry point is used by the routine at TOKENISE.
TRY_SAME_ENTRY	6F76	LD A,(IX+$00)	Has the bucket run out? Its initial letter no longer matches
	6F79	AND $1F
	6F7B	LD B,A
	6F7C	LD A,(TYPED_LETTERS)
	6F7F	CP B
	6F80	RET NZ
	6F81	PUSH HL
	6F82	LD HL,ENTRY_LETTERS	Unpack this entry's letters to ENTRY_LETTERS, by PRINT_WORD's rule for where a word ends
	6F85	LD BC,$0000
MATCH_WORD_2	6F88	LD A,(IX+$00)
	6F8B	AND $1F
	6F8D	JR Z,MATCH_WORD_3
	6F8F	LD (HL),A
	6F90	INC HL
	6F91	INC B
MATCH_WORD_3	6F92	INC IX
	6F94	INC C
	6F95	BIT 7,(IX-$01)
	6F99	JR Z,MATCH_WORD_2
	6F9B	LD A,C
	6F9C	CP $02
	6F9E	JR Z,MATCH_WORD_2
	6FA0	CP $03
	6FA2	JR NZ,MATCH_WORD_4
	6FA4	BIT 7,(IX-$02)
	6FA8	JR NZ,MATCH_WORD_2
MATCH_WORD_4	6FAA	POP HL	Keep the entry's length
	6FAB	LD A,B
	6FAC	LD (ENTRY_LENGTH),A
	6FAF	BIT 6,(IX-$01)	A synonym: step over its two-byte link to the next entry
	6FB3	RET Z
	6FB4	INC IX
	6FB6	INC IX
	6FB8	XOR A
	6FB9	RET
A linear scan, not a binary search -- which is the other half of why the list only has to be grouped by initial letter and can be loosely ordered within a group, as BLOW before BLOOD and HELP before HEART are.