This commit is contained in:
John Kerl 2016-08-11 23:57:34 -04:00
parent 8058978458
commit 9e0aebb8d0

View file

@ -1,13 +1,7 @@
================================================================
BUGFIXES
* need to ignore subproc abend (e.g. grep nonesuch) ?!?
-> + UT cases
mlr --from myfile.dat put '@v[NR][NR]=$*; emit | "grep somepattern", @*, "index1", "index2"'
...
pclose: Undefined error: 0
mlr: pclose error on "grep somepattern".
:D
================================================================
TOP OF LIST:
@ -20,67 +14,13 @@ TOP OF LIST:
! windows port?
* -f/-r field-name-spec opportunities throughout
----------------------------------------------------------------
4.4.0:
* tee and put print/dump/tee/emit >/>>/|; latter mention gzip et al.
* mlr step -a shift
* early out for unkeyed head
* sec2gmtdate is new
* sec2gmt and sec2gmtdate both leave non-numbers as-is (esp. 'NULL').
* universalize the dot operator. all this string(a).string(b) is going to get really old.
* $x=="" vs isempty($x) nonsense
* mapper nothing to mld
* copy of manpage @ web docs
----------------------------------------------------------------
POST-4.4.0:
? --imd ?
! -f/-r field-name-spec opportunities throughout
? put/tee --oxxx flags overlays ?
? mlr step -a shift --by {n}
----------------------------------------------------------------
! lemon refactor
- mdwmain w/ ifdef
- ex1/2/etc -> better names. SVNE grammar. numexpreval grammar. mlr w&w/o write/append; mlr w&w/o bracing conflict.
- clopt for no #line directives
- compile w/ as pedantic as possible but not for flex (unused input/yyunput)
- proto all
- objify all
- libify malloc_or_die etc.
- name demangles (e.g. paz wtf)
- style: typedef struct _foo_t { ... } foo_t; rid of camelcase
- substructify lemp; glorified globals at this point
- more dump routines for each stage through parser-gen & parser-operate
- replace the makefile-sedding with template names inside lempar et al.
- all this to clear the weeds & increase visibility around the memory issue
- cliopt for memset
- suspect the char/int switch
! lemon boundary issue:
! try out 2nd toy grammar
! try memset on semicolon grammar (locate the diff)
- debug:
> doesn't happen with clang -fsanitize=address
> doesn't happen with gcc -g
> nothing revealed with clang -Wall
> DSLCCOPT DSLCCDEBUG CCOPT CCDEBUG make -e
- bisect on #rules
- asan (any 2nd opinion from valgrind)
- lemon refactor
- how to do #define in .y file?
- retrieve and locally store the lemon manual
k put %ifdef in .y file
k NDEBUG
k hook in ParseTrace from cliopt
? --imd ?
----------------------------------------------------------------
? double-quote-aware (non-lite) nidx & dkvp options?