jake/SplitBit-Emulator
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Tab lists its matches in columns, and says what it can in colour

Browse Source 
Tab's listing was six instructions: print the name, print two spaces. It
was the crown of this shell for a while and looks plain next to ls.

Columns cost NOTHING EXTRA. Tab already walks its candidates twice - once
to find the answer, once to show the matches, and tabRunSources exists
because those are the same walk asked two questions. The longest match is
counted on the first, which already visits every one of them, so the
listing needs no walk of its own to know how wide a column should be.
Padding goes before a name rather than after, so a row ends on a name.

Directories are blue and the shell's own commands are green. Both are
free: Tab appends the separator itself, and a built-in is not a file at
all - which is also the only way anybody could know it will run.

A FILE IS NOT COLOURED, and that is a decision. Whether a file will run
is a read of its first block, which is what ls does and what makes ls
cost twice what it otherwise would. Here it would be worse than slow:
candidates are offered from inside a directory walk, and looking a file
up would overwrite the very fields holding the walk's own position. Doing
it safely means holding every candidate name in memory, which is a buffer
the shell would carry whether anybody pressed Tab or not - and ls is one
keystroke away. A Tab press already costs about 200,000 cycles, so the
read was not the objection.

---- And the machine could no longer build itself ----

Found by make test, not by reading. The native assembler ran out of room
for label names on cosmos.asm: 16,758 bytes against 16,384. The index was
1,341 of 1,536 in the same breath.

Which is scratch.asm's own warning happening a second time - "two
ceilings a hundred bytes apart look like one ceiling until the first is
lifted" - so both were raised, out of the seventeen kilobyte page that
file deliberately left unclaimed against exactly this. Names to 26,624
and the index to 2,048, both left about a third clear, with 1,792 bytes
still unclaimed for the same reason.

A name is thirteen bytes on average and an index entry is four, so the
arena will always be the one that speaks first. That is now written down
where the two numbers are.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E2JrLzFvuFX9fgi1LDRjrW
This commit is contained in:
Anachronaut
2026-09-06 20:18:56 -04:00
co-authored by Claude Opus 5
parent eaeb473176
commit 563bc20a75
9 changed files with 263 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Programs/CosmOS/Assembler/labels.asm
+11 -8
View File
@@ -297,18 +297,21 @@ LabEnd:
; How many labels there may be, and how many bytes of name between them.
;
; SIZED FOR THE ASSEMBLER ITSELF, which turns out to be the largest thing it is asked to
; build: 555 labels and about 6,800 bytes of name, against CosmOS's 650 and 8,081 - CosmOS is
; the bigger of the two now, and was the smaller when this was written. Running into either
; build: the assembler is about 555 labels, against CosmOS's 1,341 and 16,758 bytes of name
; measured 2026-09-06 - CosmOS is much the bigger of the two now, and was the smaller when
; this was written. Running into either
; limit says so rather than writing past the end of the table, and that is what it did. The buffers live
; above the program rather than inside it - see the scratch map in Asm.asm.
; Fifteen hundred and thirty six names, and sixteen kilobytes to hold them in. Both were
; half that, and the names ran out first: 8,081 bytes of 8,192, which is a hundred and
; eleven - and the thing that ran into it was one ordinary piece of work adding eighteen
; labels. THESE TWO MUST AGREE WITH THE SCRATCH MAP, which says where the room actually is.
; Two thousand and forty eight names, and twenty six kilobytes to hold them in. Both have
; now been doubled twice, and BOTH TIMES THE NAMES RAN OUT FIRST with the index a couple of
; hundred behind them - 8,081 of 8,192 the first time, 16,758 of 16,384 the second. A name
; is thirteen bytes on average and an index entry is four, so the arena will always be the
; one that speaks; raising it alone would buy a few hundred labels and then the other wall.
; THESE TWO MUST AGREE WITH THE SCRATCH MAP, which says where the room actually is.
LabLimit:
0x06 0x00
0x08 0x00
LabRoom:
0x40 0x00
0x68 0x00
LabNamed:
": "
Programs/CosmOS/Assembler/scratch.asm
+22 -10
View File
@@ -20,10 +20,10 @@
; bad enough; a stale number sitting next to the correction is worse, because whichever
; one a reader takes is a coin toss.
;
; 0x5000 6144 the label index, 1536 entries of four
; 0x6800 16384 the label names, packed end to end
; 0xA800 256 one block of the output file, on its way to the disk
; 0xA900 17152 free
; 0x5000 8192 the label index, 2048 entries of four
; 0x7000 26624 the label names, packed end to end
; 0xD800 256 one block of the output file, on its way to the disk
; 0xD900 1792 free
; 0xE000 1792 the vector names, 64 entries of twenty eight
; 0xE700 2048 the reader's stack, six levels of 301
; 0xEF00 368 which files have been included, sixteen names of 23
@@ -69,6 +69,17 @@
; among buffers that do not need it, because an unclaimed page is available to whichever
; one turns out to want it.
;
; ---- And one turned out to want it ----
;
; Sixteen kilobytes of names was half full when that was written and ran out on 2026-09-06,
; on an ordinary day's work on the shell: 16,758 bytes of name against 16,384. The index was
; 1,341 of 1,536 in the same breath, which is the paragraph above this one happening again -
; two ceilings, and lifting only the one that spoke would have bought two hundred labels.
;
; So both, out of the free page, and both left at about a third clear rather than at half:
; the names to 26,624 and the index to 2,048. What is left over is 1,792 bytes, kept
; unclaimed for the same reason the seventeen kilobytes were.
;
; That ends at 0xF070, with the Stack coming down from 0xFFFF above it - nearly four
; kilobytes, against the tens of bytes of CALL frames this ever nests.
;
@@ -78,16 +89,17 @@
; list moved up rather than the stack simply being asked to fit.
;
; THE TWO THINGS THAT DECIDE THESE SIZES are the largest program it will be asked to build
; and the largest one it will be asked to read. CosmOS is 475 labels and 9,564 bytes of
; output; the assembler itself is 555 labels, about 6,800 bytes of name and 11,648 of output. The
; second is bigger than the first, which is worth knowing: the hardest thing this assembles
; is not the operating system, it is itself.
; and the largest one it will be asked to read. Measured 2026-09-06: CosmOS is 1,341 labels
; and 16,758 bytes of name, and the assembler itself is far smaller. It was the other way
; round when this file was written - "the hardest thing this assembles is not the operating
; system, it is itself" - and the system has grown past it since. The numbers here are worth
; measuring again rather than believing, which is why they carry a date.
ScratchLabIndex:
0x50 0x00
ScratchLabArena:
0x68 0x00
0x70 0x00
ScratchWindow:
0xA8 0x00
0xD8 0x00
ScratchVecNames:
0xE0 0x00
ScratchSrcStack:
Programs/CosmOS/README.md
+17
View File
@@ -481,6 +481,23 @@ program; anything after it is a file.
A directory answers with a separator on the end instead of a space, which says what it is and
lets the next part be typed straight away.
**And when there is more than one, they are listed in columns**, sized to the longest of them,
the same way `ls` lays a directory out. The width comes from the screen and the longest name
is counted on the pass that finds the answer - which already visits every match, so the
listing costs no walk of its own to know how wide a column should be.
**Directories are blue and the shell's own commands are green.** Both are known for nothing:
Tab appends the separator itself, and a built-in is not a file at all - which is also why no
amount of reading the disk could have told you it will run.
**A file is not coloured, and that is a decision rather than an omission.** Whether a file
will run is a read of its first block, which is what `ls` does and what makes `ls` cost twice
what it otherwise would. Here it would be worse than slow: candidates are offered from inside
a directory walk, and looking a file up would overwrite the very fields holding the walk's own
position. Doing it safely means holding every candidate name in memory first, which is a
buffer the shell would carry whether anybody pressed Tab or not - and `ls` is one keystroke
away.
**Files are looked for in the three places the shell would look to run one**: where you are,
`/Apps` on the disk you are on, and `/Apps` on drive 0. Offering something the shell would not
find would be finishing a word into a thing that then does not work. A file that is in two of
Programs/CosmOS/Source/cosmos.asm
+208 -3
View File
@@ -1739,6 +1739,8 @@ tabIsFirst:
STA.1
SETD.1 TabBestLen
STA.1
SETD.1 TabLongest
STA.1
; A path in the word says which directory to look in, and narrows what is being matched to
; the part after the last separator.
@@ -1929,6 +1931,16 @@ tabOfferTake:
LDA.1
INCA
STA.1
; ---- How wide the widest of them is, counted on the pass that is already here ----
;
; Listing wants it and listing happens on the SECOND pass, by which time this one is over.
; Working it out there would mean a third walk of everything; working it out here costs a
; walk of one name, and only of names that matched.
CALL tabWidest
SETD.1 TabCount
LDA.1
INIB 0d1
XOR
BRQ tabOfferFirst
@@ -1939,12 +1951,62 @@ tabOfferFirst:
BRI tabOfferSkip
tabOfferPrint:
; ---- The cell the last one started, finished now ----
;
; Padding before rather than after, so a line ends on the last character of a name instead
; of on a run of spaces. The same reason ls does it that way.
SETD.1 TabAcross
LDA.1
BRA tabOfferInk
CALL tabPad
tabOfferInk:
; A directory in blue and a built-in command in green, which are the two things that can be
; known for nothing. Whether a FILE will run cannot: that is a read of its first block, and
; this is a keystroke.
SETD.1 TabKind
LDA.1
BRA tabOfferSay
DECA
BRA tabOfferBlue
INIA 0d2
BRI tabOfferInked
tabOfferBlue:
INIA 0d4
BRI tabOfferInked
tabOfferSay:
RSTA
tabOfferInked:
OUTA 0x06
SETD.1 TabStart
LDD.0.1
CALL printString
INIA 0x20
OUTA 0x00
OUTA 0x00 ; Two spaces between them, and A already holds one.
RSTA
OUTA 0x06
; How much of the cell it took, for whoever finishes it, and one more across the line.
SETD.1 TabStart
LDD.0.1
CALL tabTextLength
MVQA
SETD.1 TabUsed
STA.1
SETD.1 TabAcross
LDA.1
INCA
STA.1
SETD.1 TabColumns
LDB.1
CCF
SUB
BNQ tabOfferSkip
RSTA
SETD.1 TabAcross
STA.1
CALL newLine
tabOfferSkip:
; Past the end of this candidate, from wherever the comparison stopped.
@@ -2258,6 +2320,9 @@ tabEntry:
; same question the same way round.
; A directory, and the separator goes on the end of it.
INIA 0d1
SETD.1 TabKind
STA.1
SETD.3 TabCandidate
CALL tabEnd
INIA 0x2F
@@ -2267,6 +2332,10 @@ tabEntry:
STA.3
tabEntryFile:
RSTA
SETD.1 TabKind
STA.1
; A file, offered under the name it actually has - so nothing happens to it here at all.
; A first word used to be offered with ".sbx" taken off, and anything without that ending
; not offered, which was right while the extension was what made a file reachable by name.
@@ -2323,11 +2392,73 @@ tabNameNext:
BRA tabNamesDone
DECA
STA.1
INIA 0d2
SETD.1 TabKind
STA.1 ; A command of the shell's own, which no disk read could say.
CALL tabOffer
BRI tabNameNext
tabNamesDone:
RET
; DP0 names a string. Q is how long it is. Q rather than A, because a RET puts A back - which
; is the oldest trap on this machine and cost an afternoon in ls.
tabTextLength:
RSTA
tabTextNext:
LDB.0
BRB tabTextDone
INCA
INCD.0
BRI tabTextNext
tabTextDone:
RSTB
CCF
ADD
RET
; The candidate at TabStart against the longest seen so far, called on the pass that finds the
; answer - so listing costs no walk of its own to know how wide a column should be.
tabWidest:
SETD.1 TabStart
LDD.0.1
CALL tabTextLength
MVQA
SETD.1 TabLongest
LDB.1
CCF
SUB
BRC tabWidestKept ; Borrowed, so this one is not the longest.
STA.1
tabWidestKept:
RET
; Fills out the cell the last name started.
tabPad:
SETD.1 TabCell
LDA.1
SETD.1 TabUsed
LDB.1
CCF
SUB
BRC tabPadOne ; Wider than its cell, so one space keeps them apart.
MVQA
BRA tabPadOne
SETD.1 TabPadLeft
STA.1
tabPadNext:
INIA 0x20
OUTA 0x00
SETD.1 TabPadLeft
LDA.1
DECA
STA.1
BNA tabPadNext
RET
tabPadOne:
INIA 0x20
OUTA 0x00
RET
; ---- Showing them when they cannot be narrowed further ----
;
; The line is reprinted underneath afterwards, and where it now BEGINS is asked rather than
@@ -2335,6 +2466,52 @@ tabNamesDone:
; editor's idea of where things are hangs off those two registers.
tabList:
CALL newLine
; ---- How many of them fit across ----
;
; The longest was counted on the pass that found the answer, so this is arithmetic and not
; another walk. Two spaces between columns, and the screen says how wide it is - the same
; sum ls does, for the same reason and with the same answer.
SETD.1 TabLongest
LDA.1
INCA
INCA
SETD.1 TabCell
STA.1
INA 0x32
SETD.1 TabRoom
STA.1
RSTA
SETD.1 TabColumns
STA.1
tabFitting:
SETD.1 TabRoom
LDA.1
SETD.1 TabCell
LDB.1
CCF
SUB
BRC tabFitted
MVQA
SETD.1 TabRoom
STA.1
SETD.1 TabColumns
LDA.1
INCA
STA.1
BRI tabFitting
tabFitted:
; A candidate wider than the screen still gets a column to itself.
SETD.1 TabColumns
LDA.1
BNA tabListing
INIA 0d1
STA.1
tabListing:
RSTA
SETD.1 TabAcross
STA.1
INIA 0x01
SETD.1 TabShowing
STA.1
@@ -2342,7 +2519,14 @@ tabList:
RSTA
SETD.1 TabShowing
STA.1
; A row with something in it needs ending. One that filled its last column does not, and a
; blank line under a listing looks like a listing with a gap in it.
SETD.1 TabAcross
LDA.1
BRA tabListed
CALL newLine
tabListed:
CALL sayPrompt
INA 0x03
@@ -8728,6 +8912,27 @@ TabCopyAt:
0x00
TabFirst:
0x00
; ---- Laying the matches out in columns ----
;
; What kind of thing the candidate being offered is, how wide the widest of them was, and how
; far along a row the listing has got. The kind is set where the candidate comes FROM, since
; that is the only place that knows: 0 a file, 1 a directory, 2 one of the shell's own words.
TabKind:
0x00
TabLongest:
0x00
TabCell:
0x00
TabColumns:
0x00
TabRoom:
0x00
TabAcross:
0x00
TabUsed:
0x00
TabPadLeft:
0x00
; Whether the walk is looking for the answer or showing what the answers were.
TabShowing:
0x00