1
1
mirror of https://github.com/kanaka/mal.git synced 2024-11-11 00:52:44 +03:00
mal/basic/step8_macros.in.bas

593 lines
14 KiB
QBasic
Raw Normal View History

GOTO MAIN
REM $INCLUDE: 'mem.in.bas'
REM $INCLUDE: 'types.in.bas'
REM $INCLUDE: 'readline.in.bas'
REM $INCLUDE: 'reader.in.bas'
REM $INCLUDE: 'printer.in.bas'
REM $INCLUDE: 'env.in.bas'
REM $INCLUDE: 'core.in.bas'
REM $INCLUDE: 'debug.in.bas'
REM READ(A$) -> R
MAL_READ:
GOSUB READ_STR
RETURN
REM QUASIQUOTE(A) -> R
SUB QUASIQUOTE
REM pair?
GOSUB TYPE_A
IF T<6 OR T>7 THEN GOTO QQ_QUOTE
IF (Z%(A+1)=0) THEN GOTO QQ_QUOTE
GOTO QQ_UNQUOTE
QQ_QUOTE:
REM ['quote, ast]
B$="quote":T=5:GOSUB STRING
B=R:A=A:GOSUB LIST2
AY=B:GOSUB RELEASE
GOTO QQ_DONE
QQ_UNQUOTE:
R=Z%(A+2)
IF (Z%(R)AND 31)<>5 THEN GOTO QQ_SPLICE_UNQUOTE
IF S$(Z%(R+1))<>"unquote" THEN GOTO QQ_SPLICE_UNQUOTE
REM [ast[1]]
R=Z%(Z%(A+1)+2)
GOSUB INC_REF_R
GOTO QQ_DONE
QQ_SPLICE_UNQUOTE:
GOSUB PUSH_A
REM rest of cases call quasiquote on ast[1..]
A=Z%(A+1):CALL QUASIQUOTE
W=R
GOSUB POP_A
REM set A to ast[0] for last two cases
A=Z%(A+2)
REM pair?
GOSUB TYPE_A
IF T<6 OR T>7 THEN GOTO QQ_DEFAULT
IF (Z%(A+1)=0) THEN GOTO QQ_DEFAULT
B=Z%(A+2)
IF (Z%(B)AND 31)<>5 THEN GOTO QQ_DEFAULT
IF S$(Z%(B+1))<>"splice-unquote" THEN QQ_DEFAULT
REM ['concat, ast[0][1], quasiquote(ast[1..])]
B=Z%(Z%(A+1)+2)
B$="concat":T=5:GOSUB STRING:C=R
A=W:GOSUB LIST3
REM release inner quasiquoted since outer list takes ownership
AY=A:GOSUB RELEASE
AY=C:GOSUB RELEASE
GOTO QQ_DONE
QQ_DEFAULT:
REM ['cons, quasiquote(ast[0]), quasiquote(ast[1..])]
Q=W:GOSUB PUSH_Q
REM A set above to ast[0]
CALL QUASIQUOTE
B=R
GOSUB POP_Q:W=Q
B$="cons":T=5:GOSUB STRING:C=R
A=W:GOSUB LIST3
REM release inner quasiquoted since outer list takes ownership
AY=A:GOSUB RELEASE
AY=B:GOSUB RELEASE
AY=C:GOSUB RELEASE
QQ_DONE:
END SUB
REM MACROEXPAND(A, E) -> A:
SUB MACROEXPAND
GOSUB PUSH_A
MACROEXPAND_LOOP:
REM list?
GOSUB TYPE_A
IF T<>6 THEN GOTO MACROEXPAND_DONE
REM non-empty?
IF Z%(A+1)=0 THEN GOTO MACROEXPAND_DONE
B=Z%(A+2)
REM symbol? in first position
IF (Z%(B)AND 31)<>5 THEN GOTO MACROEXPAND_DONE
REM defined in environment?
K=B:CALL ENV_FIND
IF R=-1 THEN GOTO MACROEXPAND_DONE
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
B=R4
REM macro?
IF (Z%(B)AND 31)<>11 THEN GOTO MACROEXPAND_DONE
F=B:AR=Z%(A+1):CALL APPLY
A=R
GOSUB PEEK_Q:AY=Q
REM if previous A was not the first A into macroexpand (i.e. an
REM intermediate form) then free it
IF A<>AY THEN GOSUB PEND_A_LV
IF ER<>-2 THEN GOTO MACROEXPAND_DONE
GOTO MACROEXPAND_LOOP
MACROEXPAND_DONE:
GOSUB POP_Q: REM pop original A
END SUB
REM EVAL_AST(A, E) -> R
SUB EVAL_AST
REM push A and E on the stack
Q=E:GOSUB PUSH_Q
GOSUB PUSH_A
IF ER<>-2 THEN GOTO EVAL_AST_RETURN
GOSUB TYPE_A
IF T=5 THEN GOTO EVAL_AST_SYMBOL
IF T>=6 AND T<=8 THEN GOTO EVAL_AST_SEQ
REM scalar: deref to actual value and inc ref cnt
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
R=A
GOSUB INC_REF_R
GOTO EVAL_AST_RETURN
EVAL_AST_SYMBOL:
K=A:GOTO ENV_GET
ENV_GET_RETURN:
GOTO EVAL_AST_RETURN
EVAL_AST_SEQ:
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM setup the stack for the loop
GOSUB MAP_LOOP_START
EVAL_AST_SEQ_LOOP:
REM check if we are done evaluating the source sequence
IF Z%(A+1)=0 THEN GOTO EVAL_AST_SEQ_LOOP_DONE
REM if we are returning to DO, then skip last element
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM The EVAL_DO call to EVAL_AST must be call #2 for EVAL_AST to
REM return early and for TCO to work
Q=5:GOSUB PEEK_Q_Q
IF Q=2 AND Z%(Z%(A+1)+1)=0 THEN GOTO EVAL_AST_SEQ_LOOP_DONE
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM call EVAL for each entry
GOSUB PUSH_A
IF T<>8 THEN A=Z%(A+2)
IF T=8 THEN A=Z%(A+3)
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
Q=T:GOSUB PUSH_Q: REM push/save type
CALL EVAL
GOSUB POP_Q:T=Q: REM pop/restore type
GOSUB POP_A
M=R
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM if error, release the unattached element
REM TODO: is R=0 correct?
IF ER<>-2 THEN AY=R:GOSUB RELEASE:R=0:GOTO EVAL_AST_SEQ_LOOP_DONE
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM for hash-maps, copy the key (inc ref since we are going to
REM release it below)
IF T=8 THEN N=M:M=Z%(A+2):Z%(M)=Z%(M)+32
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM update the return sequence structure
REM release N (and M if T=8) since seq takes full ownership
C=1:GOSUB MAP_LOOP_UPDATE
REM process the next sequence entry from source list
A=Z%(A+1)
GOTO EVAL_AST_SEQ_LOOP
EVAL_AST_SEQ_LOOP_DONE:
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM cleanup stack and get return value
GOSUB MAP_LOOP_DONE
GOTO EVAL_AST_RETURN
EVAL_AST_RETURN:
REM pop A and E off the stack
GOSUB POP_A
GOSUB POP_Q:E=Q
END SUB
REM EVAL(A, E) -> R
SUB EVAL
LV=LV+1: REM track basic return stack level
REM push A and E on the stack
Q=E:GOSUB PUSH_Q
GOSUB PUSH_A
REM PRINT "EVAL A:"+STR$(A)+",X:"+STR$(X)+",LV:"+STR$(LV)+",FRE:"+STR$(FRE(0))
EVAL_TCO_RECUR:
IF ER<>-2 THEN GOTO EVAL_RETURN
REM AZ=A:B=1:GOSUB PR_STR
REM PRINT "EVAL: "+R$+" [A:"+STR$(A)+", LV:"+STR$(LV)+"]"
GOSUB LIST_Q
IF R THEN GOTO APPLY_LIST
EVAL_NOT_LIST:
REM ELSE
CALL EVAL_AST
GOTO EVAL_RETURN
APPLY_LIST:
CALL MACROEXPAND
GOSUB LIST_Q
IF R<>1 THEN GOTO EVAL_NOT_LIST
GOSUB EMPTY_Q
IF R THEN R=A:GOSUB INC_REF_R:GOTO EVAL_RETURN
A0=Z%(A+2)
REM get symbol in A$
IF (Z%(A0)AND 31)<>5 THEN A$=""
IF (Z%(A0)AND 31)=5 THEN A$=S$(Z%(A0+1))
IF A$="def!" THEN GOTO EVAL_DEF
IF A$="let*" THEN GOTO EVAL_LET
IF A$="quote" THEN GOTO EVAL_QUOTE
IF A$="quasiquote" THEN GOTO EVAL_QUASIQUOTE
IF A$="defmacro!" THEN GOTO EVAL_DEFMACRO
IF A$="macroexpand" THEN GOTO EVAL_MACROEXPAND
IF A$="do" THEN GOTO EVAL_DO
IF A$="if" THEN GOTO EVAL_IF
IF A$="fn*" THEN GOTO EVAL_FN
GOTO EVAL_INVOKE
EVAL_GET_A3:
A3=Z%(Z%(Z%(Z%(A+1)+1)+1)+2)
EVAL_GET_A2:
A2=Z%(Z%(Z%(A+1)+1)+2)
EVAL_GET_A1:
A1=Z%(Z%(A+1)+2)
RETURN
EVAL_DEF:
REM PRINT "def!"
GOSUB EVAL_GET_A2: REM set A1 and A2
Q=A1:GOSUB PUSH_Q
A=A2:CALL EVAL: REM eval a2
GOSUB POP_Q:A1=Q
IF ER<>-2 THEN GOTO EVAL_RETURN
REM set a1 in env to a2
K=A1:C=R:GOSUB ENV_SET
GOTO EVAL_RETURN
EVAL_LET:
REM PRINT "let*"
GOSUB EVAL_GET_A2: REM set A1 and A2
Q=A2:GOSUB PUSH_Q: REM push/save A2
Q=E:GOSUB PUSH_Q: REM push env for for later release
REM create new environment with outer as current environment
C=E:GOSUB ENV_NEW
E=R
EVAL_LET_LOOP:
IF Z%(A1+1)=0 THEN GOTO EVAL_LET_LOOP_DONE
Q=A1:GOSUB PUSH_Q: REM push A1
REM eval current A1 odd element
A=Z%(Z%(A1+1)+2):CALL EVAL
GOSUB POP_Q:A1=Q: REM pop A1
IF ER<>-2 THEN GOTO EVAL_LET_LOOP_DONE
REM set key/value in the environment
K=Z%(A1+2):C=R:GOSUB ENV_SET
AY=R:GOSUB RELEASE: REM release our use, ENV_SET took ownership
REM skip to the next pair of A1 elements
A1=Z%(Z%(A1+1)+1)
GOTO EVAL_LET_LOOP
EVAL_LET_LOOP_DONE:
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
GOSUB POP_Q:AY=Q: REM pop previous env
REM release previous environment if not the current EVAL env
GOSUB PEEK_Q_2
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
IF AY<>Q THEN GOSUB RELEASE
GOSUB POP_Q:A2=Q: REM pop A2
A=A2:GOTO EVAL_TCO_RECUR: REM TCO loop
EVAL_DO:
A=Z%(A+1): REM rest
GOSUB PUSH_A: REM push/save A
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM this must be EVAL_AST call #2 for EVAL_AST to return early
REM and for TCO to work
CALL EVAL_AST
REM cleanup
AY=R: REM get eval'd list for release
GOSUB POP_A: REM pop/restore original A for LAST
GOSUB LAST: REM get last element for return
A=R: REM new recur AST
REM cleanup
GOSUB RELEASE: REM release eval'd list
AY=A:GOSUB RELEASE: REM release LAST value (not sure why)
GOTO EVAL_TCO_RECUR: REM TCO loop
EVAL_QUOTE:
R=Z%(Z%(A+1)+2)
GOSUB INC_REF_R
GOTO EVAL_RETURN
EVAL_QUASIQUOTE:
R=Z%(Z%(A+1)+2)
A=R:CALL QUASIQUOTE
A=R
REM add quasiquote result to pending release queue to free when
REM next lower EVAL level returns (LV)
GOSUB PEND_A_LV
GOTO EVAL_TCO_RECUR: REM TCO loop
EVAL_DEFMACRO:
REM PRINT "defmacro!"
GOSUB EVAL_GET_A2: REM set A1 and A2
Q=A1:GOSUB PUSH_Q: REM push A1
A=A2:CALL EVAL: REM eval A2
GOSUB POP_Q:A1=Q: REM pop A1
REM change function to macro
Z%(R)=Z%(R)+1
REM set A1 in env to A2
K=A1:C=R:GOSUB ENV_SET
GOTO EVAL_RETURN
EVAL_MACROEXPAND:
REM PRINT "macroexpand"
R=Z%(Z%(A+1)+2)
A=R:CALL MACROEXPAND
R=A
REM since we are returning it unevaluated, inc the ref cnt
GOSUB INC_REF_R
GOTO EVAL_RETURN
EVAL_IF:
GOSUB EVAL_GET_A1: REM set A1
GOSUB PUSH_A: REM push/save A
A=A1:CALL EVAL
GOSUB POP_A: REM pop/restore A
IF (R=0) OR (R=2) THEN GOTO EVAL_IF_FALSE
EVAL_IF_TRUE:
AY=R:GOSUB RELEASE
GOSUB EVAL_GET_A2: REM set A1 and A2 after EVAL
A=A2:GOTO EVAL_TCO_RECUR: REM TCO loop
EVAL_IF_FALSE:
AY=R:GOSUB RELEASE
REM if no false case (A3), return nil
GOSUB COUNT
IF R<4 THEN R=0:GOSUB INC_REF_R:GOTO EVAL_RETURN
GOSUB EVAL_GET_A3: REM set A1 - A3 after EVAL
A=A3:GOTO EVAL_TCO_RECUR: REM TCO loop
EVAL_FN:
GOSUB EVAL_GET_A2: REM set A1 and A2
T=10:L=A2:M=A1:N=E:GOSUB ALLOC: REM mal function
GOTO EVAL_RETURN
EVAL_INVOKE:
CALL EVAL_AST
REM if error, return f/args for release by caller
IF ER<>-2 THEN GOTO EVAL_RETURN
REM push f/args for release after call
GOSUB PUSH_R
AR=Z%(R+1): REM rest
F=Z%(R+2)
REM if metadata, get the actual object
GOSUB TYPE_F
IF T=14 THEN F=Z%(F+1):GOSUB TYPE_F
ON T-8 GOTO EVAL_DO_FUNCTION,EVAL_DO_MAL_FUNCTION,EVAL_DO_MAL_FUNCTION
REM if error, pop and return f/args for release by caller
GOSUB POP_R
ER=-1:E$="apply of non-function":GOTO EVAL_RETURN
EVAL_DO_FUNCTION:
REM regular function
IF Z%(F+1)<60 THEN GOSUB DO_FUNCTION:GOTO EVAL_DO_FUNCTION_SKIP
REM for recur functions (apply, map, swap!), use GOTO
IF Z%(F+1)>60 THEN CALL DO_TCO_FUNCTION
EVAL_DO_FUNCTION_SKIP:
REM pop and release f/args
GOSUB POP_Q:AY=Q
GOSUB RELEASE
GOTO EVAL_RETURN
EVAL_DO_MAL_FUNCTION:
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
Q=E:GOSUB PUSH_Q: REM save the current environment for release
REM create new environ using env and params stored in function
C=Z%(F+3):A=Z%(F+2):B=AR:GOSUB ENV_NEW_BINDS
REM release previous env if it is not the top one on the
REM stack (X%(X-2)) because our new env refers to it and
REM we no longer need to track it (since we are TCO recurring)
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
GOSUB POP_Q:AY=Q
GOSUB PEEK_Q_2
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
IF AY<>Q THEN GOSUB RELEASE
REM claim the AST before releasing the list containing it
A=Z%(F+1):Z%(A)=Z%(A)+32
REM add AST to pending release queue to free as soon as EVAL
REM actually returns (LV+1)
LV=LV+1:GOSUB PEND_A_LV:LV=LV-1
REM pop and release f/args
GOSUB POP_Q:AY=Q
GOSUB RELEASE
REM A set above
E=R:GOTO EVAL_TCO_RECUR: REM TCO loop
EVAL_RETURN:
REM AZ=R: B=1: GOSUB PR_STR
REM PRINT "EVAL_RETURN R: ["+R$+"] ("+STR$(R)+"), LV:"+STR$(LV)+",ER:"+STR$(ER)
REM release environment if not the top one on the stack
GOSUB PEEK_Q_1
IF E<>Q THEN AY=E:GOSUB RELEASE
LV=LV-1: REM track basic return stack level
REM release everything we couldn't release earlier
GOSUB RELEASE_PEND
REM trigger GC
#cbm T=FRE(0)
#qbasic T=0
REM pop A and E off the stack
GOSUB POP_A
GOSUB POP_Q:E=Q
END SUB
REM PRINT(A) -> R$
MAL_PRINT:
AZ=A:B=1:GOSUB PR_STR
RETURN
REM RE(A$) -> R
REM Assume D has repl_env
REM caller must release result
RE:
R1=-1
GOSUB MAL_READ
R1=R
IF ER<>-2 THEN GOTO RE_DONE
A=R:E=D:CALL EVAL
RE_DONE:
REM Release memory from MAL_READ
AY=R1:GOSUB RELEASE
RETURN: REM caller must release result of EVAL
REM REP(A$) -> R$
REM Assume D has repl_env
SUB REP
R2=-1
GOSUB RE
R2=R
IF ER<>-2 THEN GOTO REP_DONE
A=R:GOSUB MAL_PRINT
REP_DONE:
REM Release memory from MAL_READ and EVAL
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
AY=R2:GOSUB RELEASE
END SUB
REM MAIN program
MAIN:
GOSUB INIT_MEMORY
LV=0
REM create repl_env
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
C=0:GOSUB ENV_NEW:D=R
REM core.EXT: defined in Basic
E=D:GOSUB INIT_CORE_NS: REM set core functions in repl_env
ZT=ZI: REM top of memory after base repl_env
REM core.mal: defined using the language itself
A$="(def! not (fn* (a) (if a false true)))"
GOSUB RE:AY=R:GOSUB RELEASE
A$="(def! load-file (fn* (f) (eval (read-file f))))"
GOSUB RE:AY=R:GOSUB RELEASE
A$="(defmacro! cond (fn* (& xs) (if (> (count xs) 0) (list 'if (first xs)"
A$=A$+" (if (> (count xs) 1) (nth xs 1) (throw "+CHR$(34)+"odd number of"
A$=A$+" forms to cond"+CHR$(34)+")) (cons 'cond (rest (rest xs)))))))"
GOSUB RE:AY=R:GOSUB RELEASE
A$="(defmacro! or (fn* (& xs) (if (empty? xs) nil (if (= 1 (count xs)) (first xs)"
A$=A$+" `(let* (or_FIXME ~(first xs)) (if or_FIXME or_FIXME (or ~@(rest xs))))))))"
GOSUB RE:AY=R:GOSUB RELEASE
REM load the args file
A$="(def! -*ARGS*- (load-file "+CHR$(34)+".args.mal"+CHR$(34)+"))"
GOSUB RE:AY=R:GOSUB RELEASE
IF ER>-2 THEN GOSUB PRINT_ERROR:END
REM set the argument list
A$="(def! *ARGV* (rest -*ARGS*-))"
GOSUB RE:AY=R:GOSUB RELEASE
REM get the first argument
A$="(first -*ARGS*-)"
GOSUB RE
REM no arguments, start REPL loop
IF R<16 THEN GOTO REPL_LOOP
REM if there is an argument, then run it as a program
RUN_PROG:
REM free up first arg because we get it again
AY=R:GOSUB RELEASE
REM run a single mal program and exit
A$="(load-file (first -*ARGS*-))"
GOSUB RE
IF ER<>-2 THEN GOSUB PRINT_ERROR
GOTO QUIT
REPL_LOOP:
A$="user> ":GOSUB READLINE: REM call input parser
IF EZ=1 THEN GOTO QUIT
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
IF R$="" THEN GOTO REPL_LOOP
A$=R$:CALL REP: REM call REP
IF ER<>-2 THEN GOSUB PRINT_ERROR:GOTO REPL_LOOP
PRINT R$
GOTO REPL_LOOP
QUIT:
Basic: refactor of hashmaps, map loops, remove derefs. - Alternate memory layout of hash-maps: Instead of hash-maps being an alternating sequence of keys and values, combine the key/values into a single entry. In other words, switch from this: 8 -> next Z% index (0 for last) 14 key/value (alternating) To this: 8 -> next Z% index (0 for last) key value This requires refactoring of the sequence reader, EVAL_AST and especially DO_HASHMAP and DO_KEY_VALS. So that leads to the next big refactoring: - Change mapping/lapping constructs to share code: Several pieces of mapping/looping code have a common structure, so this moves that common structure to types.in.bas: MAP_LOOP_START, MAP_LOOP_UPDATE, MAP_LOOP_DONE. Then use this code in: - EVAL_AST - READ_SEQ_* - DO_MAP - DO_HASH_MAP - DO_KEYS_VALS This also fixes the issue that several of these looping constructs were creating new empty sequence entries instead of using the common ones at the beginning of memory. - Remove the use of DEREF_*. This isn't actually needed because we no longer create structure that refer to multiple levels of type 14 references. Replace DEREF_* with VAL_* which gets the value of a particular sequence element i.e. Z%(A+1,1). All together, the above changes save over 300 bytes. Also: - Fix empty nil/false/true entries so they are treated the same as other types of data with regards to reference counting and ALLOC/RELEASE. - Add a new memory summary function in debug.in.bas that just prints out free, value count, and references for the early scalar and empty list elements. Comment out the larger one. This saves about 90 bytes.
2016-11-16 07:38:09 +03:00
REM GOSUB PR_MEMORY_SUMMARY_SMALL
#cbm END
#qbasic SYSTEM
PRINT_ERROR:
PRINT "Error: "+E$
ER=-2:E$=""
RETURN