正则表达式匹配模式直到下一次出现

Question

I have following data:

2018-03-20 23:28:47 INFO This is an info sample(can be multiline with new line characters)
2018-03-20 23:28:47 INFO This is an info sample(can be multiline with new line characters)
2018-03-20 23:28:47 DEBUG This is a debug sample(can be multiline with new line characters) {
  'x':1,
  'y':2,
  'z':3,
  'w':4
}
2018-03-20 23:28:47 INFO This is an info sample(can be multiline with new line characters)
2018-03-20 23:28:47 DEBUG This is a debug sample(can be multiline with new line characters){
  'a':5,
  'b':6,
  'c':7,
  'd':8
}

I've to extract all DEBUG statements and for that I am using this regex (\\d{4}\\-\\d{2}\\-\\d{2}\\ \\d{2}\\:\\d{2}\\:\\d{2}\\ DEBUG(.|\\n|\\r)*?)(?=\\d{4}\\-\\d{2}\\-\\d{2}\\ \\d{2}\\:\\d{2}\\:\\d{2}) but it is omitting the last DEBUG statement. What should be the regex to obtain following output?

2018-03-20 23:28:47 DEBUG This is a debug sample(can be multiline with new line characters) {
  'x':1,
  'y':2,
  'z':3,
  'w':4
}
2018-03-20 23:28:47 DEBUG This is a debug sample(can be multiline with new line characters){
  'a':5,
  'b':6,
  'c':7,
  'd':8
}

Answer 1

I suggest:

• Anchor the matches at the start of the line to make it safer (by using (?m) )
• Fix the current issue by adding an alternative with the very end of the string \\Z (same as Ken suggests in the comments)
• Replace a very inefficient (.|\\r|\\n)*? pattern with .*? and adding a DOTALL modifier (?s)

The whole fix will look like

(?sm)^\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2} DEBUG\s*(.*?)(?=[\r\n]+\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}|\Z)

See the regex demo .

Details

• (?sm) - DOTALL and MULTILINE options on
• ^ - start of a line
• \\d{4}-\\d{2}-\\d{2} \\d{2}:\\d{2}:\\d{2} - a timestamp like pattern
• DEBUG - a literal substring
• \\s* - 0+ whitespaces
• (.*?) - Group 1: any 0+ chars, as few as possible, up to but excluding
• (?=[\\r\\n]+\\d{4}-\\d{2}-\\d{2} \\d{2}:\\d{2}:\\d{2}|\\Z) - a positive lookahead that requires either
  • [\\r\\n]+\\d{4}-\\d{2}-\\d{2} \\d{2}:\\d{2}:\\d{2} - one or more CR or LF symbol(s) followed with a timestamp like pattern
  • | - or
  • \\Z - the very end of the string

Answer 2

If you are sure that all the paragraphs with DEBUG will end with } , you can use:

r"(.*DEBUG[\s\S]*?\})"

If DEBUG may or may not have {} , the following regex should do the trick:

r"(.*DEBUG.*(?!=\{|\n))(\{[\s\S]*?\})?"