正则表达式：字符串开头不符合预期

Question

I am parsing (by regex) a .map file generated by a linker for ARM. I have extracted pretty much everything but this section is resisting.

Here is an excerpt of the part I want to parse

COMMON         0x20002b18        0x1 ./2_Programa/source/board.o
               0x20002b18                BOARD_ctx
COMMON         0x20002b19       0x87 ./2_Programa/source/interface_objects.o
               0x20002b19                GLB_appIntObjPropChangeFlags
               0x20002b1a                GLB_aioBLCommand
               0x20002b65                GLB_aioDateTime
COMMON         0x20002ba0       0x31 ./2_Programa/source/objects.o
               0x20002ba0                GLB_goFlags
*fill*         0x20002bd1        0x3 

and this is my best regex attempt:

^ COMMON\s+(0x\S+)\s+(0x\S+).*(?:\s+(0x\S+)\s+(\S+)[\r\n])*(?:\s+\*fill\*\s+0x\S+\s+(0x\S+))?

The result can be checked here . The result I get, only matches the last line of the block (I consider a block when it starts with COMMON ).

What I need to extract is something similar to this:

[{
    'name': 'GLB_appIntObjPropChangeFlags',
    'size': 0x01,
    'path': './2_Programa/source/interface_objects.o',
    'origin': 0x20002b19
},
{
    'name': 'GLB_aioBLCommand',
    'size': 0x87,
    'path': './2_Programa/source/interface_objects.o',
    'origin': 0x20002b1a
},
...
]

My main problem here is that I am not able to separate the first line

COMMON 0x20002b19 0x87 ./2_Programa/source/interface_objects.o`

from the others related to it

        0x20002b19                GLB_appIntObjPropChangeFlags
        0x20002b1a                GLB_aioBLCommand
        0x20002b65                GLB_aioDateTime

Could anyone give some hints to face this off?

UPDATE

What I would like to do is to split all blocks (those that start with COMMON ) into two parts. Group 1:

COMMON 0x20002b19 0x87 ./2_Programa/source/interface_objects.o`

and Group2:

        0x20002b19                GLB_appIntObjPropChangeFlags
        0x20002b1a                GLB_aioBLCommand
        0x20002b65                GLB_aioDateTime

Then, I could regex each group separately:

Regex for Group 1:

^ COMMON\s+(0x\S+)\s+(\S+)\s+(\S+)

and this other for Group 2 (setting multi line flag):

^\s+(0x\S+)\s+(\S+)

As a result I will get three groups from first regex and other six (2 per line per 3 lines) which could easily converted in a list of dict s as I showed above.

Answer 1

Brief

Realistically, you should grab each COMMON block as mentioned in the comments under your question by Wiktor Stribiżew . Link to Wiktor's regex here . Regex does not have the ability to loop over a subquery (that's not its purpose).

Impractically, you can use this regex to grab each COMMON section and its following blocks, and then map it.

---

Code

See regex in use here

(?:COMMON\s+0x[0-9a-f]+\s+(0x[0-9a-f]+)\s+(\S+)|\s*(0x[0-9a-f]+)\s+(\S+))(?=\s*[\r\n])

---

Explanation

• COMMON\\s+0x[0-9a-f]+\\s+(0x[0-9a-f]+)\\s+(\\S+) Option 1
  • COMMON\\s+0x[0-9a-f]+\\s+
    • COMMON The characters COMMON literally
    • \\s+ One or more whitespace characters
    • 0x These characters 0x literally
    • [0-9a-f]+ One or more of the characters in the set 0-9a-f
    • \\s+ One or more whitespace characters \\s+
  • (0x[0-9a-f]+) Capture the following into capture group 1
    • 0x These characters 0x literally
    • [0-9a-f]+ One or more of the characters in the set 0-9a-f
  • \\s+ One or more whitespace characters
  • (\\S+) Capture one or more non-whitespace characters into capture group 2
• \\s*(0x[0-9a-f]+)\\s+(\\S+) Option 2
  • \\s* Any number of whitespace characters
  • (0x[0-9a-f]+) Capture the following into capture group 3
    • 0x These characters 0x literally
    • [0-9a-f]+ One or more of the characters in the set 0-9a-f
  • \\s+ One or more whitespace characters
  • (\\S+) Capture one or more non-whitespace characters into capture group 4
• (?=\\s*[\\r\\n]) Ensure what follows is any number of whitespace characters, followed by a newline character \\r\\n

---

Usage

Based on the order of the matches and the groups to which they belong, you can map them to an array as you've presented.

For example (in match order).

• First set
  • Group 1 0x1
  • Group 2 ./2_Programa/source/board.o
  • Group 3 0x20002b18
  • Group 4 BOARD_ctx
• Second set
  • Group 1 0x87
  • Group 2 ./2_Programa/source/interface_objects.o
  • Group 3 0x20002b19
  • Group 4 GLB_appIntObjPropChangeFlags
  • Group 3 0x20002b1a
  • Group 4 GLB_aioBLCommand
  • Group 3 0x20002b65
  • Group 4 GLB_aioDateTime
• etc.

Always associating the last match for group 1 and group 2 to the current match for group 3 and group 4