CVS log for parser3/src/include/pa_string.h

Up to [parser3project] / parser3 / src / include

Request diff between arbitrary revisions - Display revisions graphically

Keyword substitution: kv
Default branch: MAIN
Current tag: MAIN

Copyright year updated, websites links changed to https://

feature #1273 core implementation: value::as_wint added, implemented for int/double/string, and used where applicable; vint now holds pa_wint type; format_double moved to pa_int

pa_int includes tuning, vcproj updated (related to feature #1273)

pa_int.h/pa_int.C added and all related code moved there, pa_wint and related functions added, another preparation for feature #1273

implements feature #1277
String::Body(const char *abody) now explicit
but HashString::put(const char *str, V value) added for simplicity
optimizations:
HASH_PUT_CSTR used where hash.put called several time during execution;
caching String::Body::uitoa used at all pash.put(<uint> calls.

String::Body::uitoa added for caching (related to feature #1277)

optimized String:pos versions added (without language + char *) (related to feature #1277)

Copyright year updated

faster templated pa_itoa/pa_uitoa added to work with any integer type, String::Body::Format removed

ArrayString get() null result assert also added (related to bug #1238)

In assertions enabled mode, ArrayString values are now required to be non-null (related to bug #1238)

'+' and '-' strings are no longer can be treated as number 0, exception thrown (fixes bug #1234)

copyright, authors actualized

Copyright year updated

long long -> int64_t

pa_atoi / pa_atoui / pa_atoul usage checked. 10 is now default base, no hex autodetect by default

bugfix for Windows x64: enum Language -> enum Language : size_t - required for VS2015+ to make sizeof(Languages::opt) == sizeof(CORD), will be 16 byte under x64 without it (related to feature #1198)

to be sure that our version of gc.h/cord.h is used, we now include it directly

Copyright year: 2015->2017

^reflection:tainting[$string] added (implements feature #1098)

pa_exec now supports $.stdin[$binary_file] (fixes bug #1044)

optimization: L_PASS_APPENDED removed as no longer required (related to feature #1091)

cleanup: string::split limit argument removed as no longer required, pos_after no longer reference (related to feature #11)

added StringSplitHelper as friend (related to feature #11)

body and langs now protected, not private to allow inheritance (related to feature #11)

whitespace

initial implementation for symbols cache (related to feature #1026)

Copyright year updated

explicit String::Body(char) calls removed; String::Body(String::C) constructor added (related to bug #957)

String(char*, Lang, length) removed, String(C(), Lang) used instead (related to bug #957)

Now char * is converted to CORD only using AS_CORD() to check for empty cord bug (fixes bug #957)

pa_atoul added for unsigned long long int support

warning war (-pedantic flag)

removed most of the clang compiler warnings (closes issue #1018)

String::Body comparation operators with char * added (related to issue #958)

warning war

no more double slashes in file path (fixes bug #872)

pa_atoui added for out of range checks, zero division in vint/vdouble check added ( fixes #832 )

- strrpbrk & rskipchars were added

 Serialization_data now added and used, VString now serialized with languages into memcached.

ident now works under Linux + ident displays filenames (closes issue #818)
Copyright updated

files upload now uses binary blocks instead of L_FILE_POST tainting. (bugfix: #128)

- method append_quoted was added

- taint[json] was added
- escaping was slightly modified

- optimisation: cacheing of cstr was added

- little hacking: for dealing with transcoding+^taint[uri] transcode_and_untaint was replaced by untaint_and_transcode

- some fixes with transcode+untaint

- optimization: CORD_str, CORD_to_const_char_star, CORD_to_char_star and CORD_substr have one more arg: length of cord

- use cached string length in Body::mid. it improves speed of match with brackets in pattern

- new language L_PARSER_CODE for auto-untaint in process was added

- some code cleanup

- cstr(params) and cstrm(params) were deleted. taint_cstr[m] and untaint_cstr[m] should be used

- optimizations and simplification of string.cstr(...)

- trim can work with utf-8 chars now

- string length cacheing was implemented

- CORD_cat optimization

- String constructor which can print formatted string was added

- type was fixed

- hash_code caching was implemented (seaches should be faster now)

- String constructors don't have that stupid true/false param which means 'tainted'. they accept lang instead.

- result optimisation
- helper_length parameter removed from string constructors

- little optimisation: somewhere length() was replaced by is_empty()

- parser is x64 friendly now

- use vregex object instead of direct calls methods from pcre lib

- not needed template removed

- String::append optimisation

- for file post the new taint language L_FILE_POST was added

- escape method was added

no message

- helper length added for mid for small optimisation

- methods for working with pos/mid for strings in utf-8 were added

- taint lang (internal) L_HTTP_COOKIE was added. will used for escaping cookies as %uXXXX

- was compilation error during build on freebsd4

- ^string.match[][] understand new option now: return number of matches but not table wit results

- bool class was added

+ ^taint/untaint[regex] << escapes these: \^$.[]|()?*+{}

2005

debugger help: String.dump() to stdout in detailed form  [were String.v() in short form]

beauty: removed outdated level 'table'

cancel: more speed, less memory: CORD_chars_block originally intended to ...
it was good that CORD_chars_block were not used: it consumes more memory [and, might be slow too]
undoing that. [and removed that func so that it would not confuse parser developer in future]

more speed, less memory: CORD_chars_block originally intended to speedup/reduce mem usage were forgotten, and were used stupid CORD_chars

bugfix: on BIGENDIAN processors space-conserving mech failed, causing
SIGSEGV/SIGBUS and SAPI::abort("unknown untaint lang#%d", (1|2|3));

new: @method[][result] means "no string output here",
implemented part of that idea: compiler throws away string literal generation code. and barks nonwhitespace chars

new: ^file::sql[[name_to_become_$.name]]{}
query result must be one row with columns:
  first: data
  second: file name
  third: content-type

new method: ^string.trim[start|both|end[;chars]]        default 'chars' -- whitespace chars
  started, only start works

2004

bugfix: ^date::create[invalid fields now produce exception.

bugfix: removed too strong a assert

static const char * const IDENT

more warnings --

fixed: string.match[g] without <'> option produced columns with NULL's, which gpf'ed at table.save time

CORD_chr does not check offset argument for validity
did that in String::Body::pos myself
http://i2/tasks/edit/?id=4577425257580789777

String.for_each bug fixed [omited one-char cases]

string::serialize/deserialize implemented
Language enum assigned meaningful letters [more convinient for debugging. read warning before adding/changing anything]

templates and anonymous unions differences on unix

string_fragments_to_cord merged to HEAD

moved tempate_gc to HEAD

2002->2003

String::first_char now not fails on empty strings
^if(-f '') now ok and in 3 other places.

String::first_char now not fails on empty strings
^if(-f '') now ok and in 3 other places.

$Date: now$

ident.C* removed

uue to separate file
introducting $message:received
^mail:send rewritten with backward comp

^string.optimize[]

string now optimized prior to .left .right .mid .pos .match when that is profitable,
and always before .replace
when $ORIGINS(1) optimization disabled
economy from not wasting mem on lots of strings which occur by lots of String:mid calls to get parts of source string between found_occurances

string.match[]['] option enables generation of $match.prematch .match .postmatch columns

killed 8 bytes from String.head

fixed string.replace [when reconstructing pieces were split by max_integral(piece.size), thus some strings to replace happen to be split into two = not replaced)

mail subject encoding taken from .content-type.charset, not .charset

string deserialize checks on broken file

string iterators fixed again, so were String::join_chain

STRING_*FOREACH_ROW
changed to stop at append_row, not at link zero
now string to string appending works

//#define DEBUG_STRING_APPENDS_VS_EXPANDS

economy: 22%

space uptimized: string::link_row removed

speed uptimized: string::is_empty

strign to string append optimiziation idea and estimates

name spelling

2002

table set & append changed splitting languages, separators now can be clean AND as-is
same to mail command line $MAIN:MAIL.progX arguments

^mail:send[
   $.charset[zzz]
addded

introducing Charset

UL_CLEAN pieces now inherit UL_OPTIMIZED_BIT at wcontext.write

removed 8 bytes from String, 2 bytes from string::chunk::row::item

removed 1+4 bytes from String

introducing estimating String::cstr_bufsize, 0

$status:db
fixed used not initialized in db_connection & db_table

String & Origin packed with #pragma pack

memory hunging
#pragma pack(1)

16 419 844
13 169 394

String size_t replace to uint in proper places [todo:Array]

debugging memory
String::fused_rows func now

[Mon Oct 29 17:55:45 2001] total 5917436/158731
[Mon Oct 29 18:11:53 2001] total 5796092/158731

debugging memory
hash key ref now

[Mon Oct 29 16:01:17 2001] total 17050496/202647
[Mon Oct 29 17:11:17 2001] total 16867948/202658

debuggging memory.
string::as_int/double now usually uses stack, not heap [200K out of 17M :)]
VString(String&) not copies reference [29M -> 17M]

hashfile pieced serialize. todo stamp

restructured string: linked pieces of same language together prior to String::replace

untaint lang origins table fixed.
got rid of empty_string, which caused errors - it hasnt exception but somebody[value.bark] tried to throw it on it's pool

added ident.awk and main.dsp splitted to main+classes+types along to .am

merged with before_xml

dom.xslt  stylesheet compiled&cached

^dom:set{<some>xml</some>} default language XML
languages+=xml

pa_directory speeded up. moved zero 'from' check to directory constructor.
string:replace

speeded up typo-html replacements. introducing Dictionary with first-char caching

only ^class:method dynamic calls allowed.
^BASE.method call disabled.
BASE element globally removed

$ORIGINS(1) output tracing mode

date roll
table calendar

started int:sql

started int:sql

introducing String::cstr_bufsize, returns just size+1 for as_is target.

#include "pa_config_includes.h"
in all headers

array debugged; adjusted

string+array made linear grows

think that all must grow lineary, not exponentialy

string fixed bug with fullchunk cmps

detected huge mem allocation: size/times malloc 27809390/368771, calloc 3232/83. would test now

^string.upper|lower[]

dox, split by not clean parts also

z

code documentation ++

untaint - 	@test optimize whitespaces for all but 'html'

untaint - 	@test mail-header

$MAIN:LOCALE

fixed vstring:as_vfile  length

sapi exec dead end. badly parsed args in apache:util.script

written but not tested exec with env for unix

exec class just compiled. moved post read to request core

uuencode. string<<

mail:send -1 text prepared

mail:send -10 just compiled

hash now not thread-safe.

z

sql quote. string untaint UL_SQL

connect&sql -1000 just compiled

string:match replace 2.  string.cstr(forced lang)

string:match replace strange matches

string:match replace prepared

string:match replace -11

string:match 0

string:match [search] -1 just compiled

string:match -10 [frame]

string::pos & cmp & piece bugs fixed.
string::split new
table:load separated from set
table:set implemented with clean \n \t searches

string::pos. about to use it in table:set/load

table:join

todo: table flip and append. string::pos

string: lsplit

string: lsplit -1

string: pos

string: pos -1

string: left right mid

#comment fixed minor bug

typo & typo-default

html-typo 2

html-typo 0
html-typo 1. noticed double default content-type prob. run cgi.cmd tomorrow first
html-typo sample. detected << problem: they become ltlt too fast

cgi: reinvent document_root and request_uri under iis. autoconf updated. introducing ap_config_includes

shifted apache branch on main trunc

z

z

file 1

file class just compiled

z

auto @brief

doxygen include/, Value

cookie class

$defautl[$content-type[$value[text/html] $charset[windows-1251]]

^taint 0

convinient string(pool, char *src, bool tainted) ctor

proper @exceptions params tainting

global handler, exceptions in parser3

'value is type' expr operator

table:set 0

^process. actual names to store param and  check_actual_numbered_params for better place diagnostics

^untaint 0

headers updated

sources header

tainting 0

^lang prepare0

just compiled lt&co

write_value write_string 0
it seems wcontext must write strings regardles of fvalue!=0

string.operator==(char*)

struck with const array.gets

bison[yacc] first time compiled.
execute=dump for now

get names 0

process text repassing

String_iterator tested

String_iterator optimized

String_iterator::skip_to
todo:optimize

started String_iterator

:pooled

Table
more like C++ style
hence lots of 'const'

Error fixed

added forgotten

Request
Error
Table

Table started
would test
template Array now

String originating

moved .h public parts to top

array cache rethought to chunk caching

Array 0

no message

uint, and added pa_types & pa_hash[forgotten]

removed templates [vc++ suxx]

templates in VC++ suxx.

pa_pool split