PCRE正则表达式，从包含具有不同分隔符和条带注释的多个语句的字符串中提取单个SQL语句

Question

i am trying to use regular expressions to extract singular sql statements from a file containing several sql statements and alternate delimiters/comments.

i am trying to match the following patterns to isolate sql statements, then after isolating an individual statement, stripping it of comments: "delimiter (del) (nonwhitespace sequence) (not (del) or comment with (del)) (del)" "(not ; ) ;"

the first pattern should allow the use of any set of characters for a delimiter

---

i tried the following to match the first pattern:

"/\s*delimiter\s+(?<d>[^\s]+)\s*;?\s*(?<qstr>(((?!--|\g{d}).)+|--[^\R]*\R)*)\g{d}\s*;?/s"

and if the first pattern fails, to match the second pattern:

"/\s*(?<qstr>(((?!--|;).)+|--[^\R]*\R)*);/s"

then if either succeeds, replace the following with empty string:

"/--[^\n\r]*(?:\n|\r)*/"

my problem is that apache crashes on preg_match when i try to search for either of the first 2 regular expressions on the following string:

"delimiter $$
create table MovieDetail
(
   imdbid varchar(32) primary key not null,
   title varchar(512),
   year int,
   rated varchar(16),
   released int,
   runtime int,
   director varchar(128),
   writer varchar(12),
   plot varchar(2048),
   imageurl varchar(512),
   rating float,
   ratingcount int,
   type varchar(64)
); $$
detect this text as a separate statement"

the first match should be

"delimiter $$
create table MovieDetail
(
   imdbid varchar(32) primary key not null,
   title varchar(512),
   year int,
   rated varchar(16),
   released int,
   runtime int,
   director varchar(128),
   writer varchar(12),
   plot varchar(2048),
   imageurl varchar(512),
   rating float,
   ratingcount int,
   type varchar(64)
); $$"

and the subpattern <qstr> should be

"create table MovieDetail
    (
       imdbid varchar(32) primary key not null,
       title varchar(512),
       year int,
       rated varchar(16),
       released int,
       runtime int,
       director varchar(128),
       writer varchar(12),
       plot varchar(2048),
       imageurl varchar(512),
       rating float,
       ratingcount int,
       type varchar(64)
    )"

The goal is to extract the first sql statement in a string containing multiple sql statements. it can then determine a new index in the string after accounting for the extracted statement and proceed to extract the next sql statement from that index. the goal is to allow my script to execute individual sql statements from a string containing multiple sql statements so that it can print the individual results for each statement (failure/success/fetched results from a query if any). delimiter is not part of sql and i need it to enable my script to define triggers or sql stored programs which contain multiple sql statements but need to be treated as one.

i tried replacing escape sequences with // like //s and //g and it still crashes just the same. i tried testing them on debugex.com and both expressions are valid. i'm using XAMPP with Apache 2.4.17 and PHP 5.6.23 (VC11 X86 32bit thread safe) + PEAR.

major update: i found out the error only occurs when running the regex expression on multi-line strings, so i'm going to try comparing the binary data of the string with that of one replacing line breaks by \\n or \\r\\n

i also realized the regex expressions above don't account for sql string-expressions, so the updated regex's are

"/\s*delimiter\s+(?<d>[^\s]+)\s*;?\s*(?<qstr>(((?!--|\g{d})[^'])+|'([^']|'')*'|--[^\R]*\R)*)\g{d}\s*;?/sA"

and

"/\s*(?<qstr>((?!--|;)[^']|'([^']|'')*'|--[^\R]*\R)*);/sA"

the crash is related to multiple whitespace characters. if i do

preg_replace("/\s+/", " ", $sqlstr);

the crash is eliminated and it functions properly. also, if i do

preg_replace("/[ ]+/", " ", preg_replace("/\R+/", "\n", $sqlstr));

the crash is also eliminated. i can use that as a work-around but i don't like it because it doesn't preserve the original string. it also might cause problems extracting subsequent sql statements since the index will correspond to the modified not the original string.

Answer 1

I found out the source of the crash was the use of nested and repeated captured subpatterns, but even beyond that, the expressions did not work. i eventually gave up and resorted to manual character processing.