ed: make scratchpad memory only

The dual scraptchpad thing doesn't work. Things become very
complicated when it's time to write that back to the file. We
overwrite our contents and end up with garbage.

apps/ed/buf.asm

@@ -1,29 +1,19 @@
 ; buf - manage line buffer
 ;
-; Lines in edited file aren't loaded in memory, their offsets is referenced to
-; in this buffer.
-;
-; About scratchpad and offsets. There are two scratchpads: file and memory.
-; The file one is the contents of the active blkdev. The second one is
-; in-memory, for edits. We differentiate between the two with a
-; "scatchpad mask". When the high bits of the offset match the mask, then we
-; know that this offset is from the scratchpad.
-;
 ; *** Consts ***
 ;
 ; Maximum number of lines allowed in the buffer.
 .equ	BUF_MAXLINES	0x800
 ; Size of our scratchpad
 .equ	BUF_PADMAXLEN	0x1000
-; Scratchpad mask (only applies on high byte)
-.equ	BUF_SCRATCHPAD_MASK	0b11110000
 
 ; *** Variables ***
 ; Number of lines currently in the buffer
 .equ	BUF_LINECNT	BUF_RAMSTART
-; List of words pointing to scratchpad offsets
+; List of pointers to strings in scratchpad
 .equ	BUF_LINES	BUF_LINECNT+2
-; Points to the end of the scratchpad
+; Points to the end of the scratchpad, that is, one byte after the last written
+; char in it.
 .equ	BUF_PADEND	BUF_LINES+BUF_MAXLINES*2
 ; The in-memory scratchpad
 .equ	BUF_PAD		BUF_PADEND+2
@@ -35,20 +25,34 @@
 ; On initialization, we read the whole contents of target blkdev and add lines
 ; as we go.
 bufInit:
-	ld	hl, BUF_PAD
-	ld	(BUF_PADEND), hl
-	ld	ix, BUF_LINES
+	ld	hl, BUF_PAD	; running pointer to end of pad
+	ld	de, BUF_PAD	; points to beginning of current line
+	ld	ix, BUF_LINES	; points to current line index
 	ld	bc, 0		; line count
 .loop:
-	call	ioTell		; --> HL
 	call	ioGetC
 	jr	nz, .loopend
-	ld	(ix), l
+	or	a		; null? hum, weird. same as LF
+	jr	z, .lineend
+	cp	0x0a
+	jr	z, .lineend
+	ld	(hl), a
+	inc	hl
+	jr	.loop
+.lineend:
+	; We've just finished reading a line, writing each char in the pad.
+	; Null terminate it.
+	xor	a
+	ld	(hl), a
+	inc	hl
+	; Now, let's register its pointer in BUF_LINES
+	ld	(ix), e
 	inc	ix
-	ld	(ix), h
+	ld	(ix), d
 	inc	ix
 	inc	bc
-	call	ioGetLine
+	ld	(BUF_PADEND), hl
+	ld	de, (BUF_PADEND)
 	jr	.loop
 .loopend:
 	ld	(BUF_LINECNT), bc
@@ -65,45 +69,20 @@ bufLineAddr:
 	pop	de
 	ret
 
-; Read line number specified in HL and loads the I/O buffer with it.
-; Like ioGetLine, sets HL to line buffer pointer.
+; Read line number specified in HL and make HL point to its contents.
 ; Sets Z on success, unset if out of bounds.
 bufGetLine:
-	push	de
+	push	de		; --> lvl 1
 	ld	de, (BUF_LINECNT)
 	call	cpHLDE
-	jr	nc, .outOfBounds	; HL > (BUF_LINECNT)
-	call	bufLineAddr
-	; HL now points to seek offset in memory
-	ld	e, (hl)
-	inc	hl
-	ld	d, (hl)
-	; DE has seek offset
-	ex	de, hl
-	pop	de
-	; is it a scratchpad offset?
-	ld	a, h
-	and	BUF_SCRATCHPAD_MASK
-	cp	BUF_SCRATCHPAD_MASK
-	jr	z, .fromScratchpad
-	; not from scratchpad
-	cp	a	; ensure Z
-	jp	ioGetLine	; preserves AF
-.fromScratchpad:
-	; remove scratchpad mask
-	ld	a, BUF_SCRATCHPAD_MASK
-	xor	0xff
-	and	h
-	ld	h, a
-	; HL is now a mask-less offset to BUF_PAD
-	push	de	; --> lvl 1
-	ld	de, BUF_PAD
-	add	hl, de
 	pop	de		; <-- lvl 1
+	jp	nc, unsetZ	; HL > (BUF_LINECNT)
+	call	bufLineAddr
+	; HL now points to an item in BUF_LINES.
+	call	intoHL
+	; Now, HL points to our contents
+	cp	a		; ensure Z
 	ret
-.outOfBounds:
-	pop	de
-	jp	unsetZ
 
 ; Given line indexes in HL and DE where HL < DE < CNT, move all lines between
 ; DE and CNT by an offset of DE-HL. Also, adjust BUF_LINECNT by DE-HL.
@@ -150,7 +129,7 @@ bufDelLines:
 ; Insert string where DE points to memory scratchpad, then insert that line
 ; at index HL, offsetting all lines by 2 bytes.
 bufInsertLine:
-	push	de	; --> lvl 1, scratchpad offset
+	push	de	; --> lvl 1, scratchpad ptr
 	push	hl	; --> lvl 2, insert index
 	; The logic below is mostly copy-pasted from bufDelLines, but with a
 	; LDDR logic (to avoid overwriting). I learned, with some pain involved,
@@ -187,22 +166,15 @@ bufInsertLine:
 	ld	(hl), d
 	ret
 
-; copy string that HL points to to scratchpad and return its seek offset, in HL.
+; copy string that HL points to to scratchpad and return its pointer in
+; scratchpad, in HL.
 bufScratchpadAdd:
 	push	de
 	ld	de, (BUF_PADEND)
 	push	de	; --> lvl 1
 	call	strcpyM
+	inc	de		; pad end is last char + 1
 	ld	(BUF_PADEND), de
 	pop	hl	; <-- lvl 1
-	; we have a memory offset in HL, but it's not what we want! we want a
-	; seek offset stamped with the "scratchpad mask"
-	ld	de, BUF_PAD
-	scf \ ccf
-	sbc	hl, de
-	ld	a, h
-	or	BUF_SCRATCHPAD_MASK
-	ld	h, a
-	; now we're good...
 	pop	de
 	ret

apps/ed/io.asm

@@ -32,7 +32,7 @@ ioInit:
 	ld	ix, IO_FILE_HDL
 	jp	fsPutC
 .blkdev:
-	.dw	.fsGetC, .fsPutC
+	.dw	.fsGetC, unsetZ
 
 ioGetC:
 	push	ix
@@ -61,34 +61,3 @@ ioTell:
 	call	_blkTell
 	pop	ix
 	ret
-
-; Given an offset HL, read the line in IO_LINE, without LF and null terminates
-; it. Make HL point to IO_LINE.
-ioGetLine:
-	push	af
-	push	de
-	push	bc
-	ld	de, 0		; limit ourselves to 16-bit for now
-	xor	a		; absolute seek
-	call	ioSeek
-	ld	hl, IO_LINE
-	ld	b, IO_MAXLEN
-.loop:
-	call	ioGetC
-	jr	nz, .loopend
-	or	a		; null? hum, weird. same as LF
-	jr	z, .loopend
-	cp	0x0a
-	jr	z, .loopend
-	ld	(hl), a
-	inc	hl
-	djnz	.loop
-.loopend:
-	; null-terminate the string
-	xor	a
-	ld	(hl), a
-	ld	hl, IO_LINE
-	pop	bc
-	pop	de
-	pop	af
-	ret

apps/ed/main.asm

@@ -15,16 +15,10 @@
 ;
 ; That doubly linked list on the z80 would use 7 bytes per line (prev, next,
 ; offset, len), which is a bit much. Moreover, there's that whole "scratchpad
-; being loaded in memory" thing that's a bit iffy. We sacrifice speed for
-; memory usage.
+; being loaded in memory" thing that's a bit iffy.
 ;
-; So here's what we do. First, we have two scratchpads. The first one is the
-; file being read itself. The second one is memory, for modifications we
-; make to the file. When reading the file, we note the offset at which it ends.
-; All offsets under this limit refer to the first scratchpad. Other offsets
-; refer to the second.
-;
-; Then, our line list is just an array of 16-bit offsets. This means that we
+; We sacrifice speed for memory usage by making that linked list into a simple
+; array of pointers to line contents in scratchpad. This means that we
 ; don't have an easy access to line length and we have to move a lot of memory
 ; around whenever we add or delete lines. Hopefully, "LDIR" will be our friend
 ; here...