如何通过正则表达式定位{}内最后逗号分隔后的子串并捕获其中的下划线？

云上老周

2026-4-29

Solution to Target Underscores in Last Comma-Separated Segment Inside Curly Braces

Got it, let's break down how to solve this problem so you can precisely target and replace those underscores.

Your current regex matches everything inside {}, but we need to narrow it down to only underscores that live in the last comma-separated segment of each {...} block. Here's a step-by-step solution:

The Target Regex

Use this regex to match exactly the underscores you need:

_(?=[^},]*})(?<=\{.*,)

How It Works

Let's break down each part to understand why this works:

_: The literal underscore we want to match.
(?=[^},]*}): A positive lookahead that ensures, from the underscore to the closing }, there are no commas. This guarantees we're in the last segment of the {...} block.
(?<=\{.*,): A positive lookbehind that ensures the underscore is inside a {...} block (we see an opening { before it) and that there's at least one comma before the underscore (so we skip underscores in earlier segments).

Example Usage with `regexp_replace`

Let's apply this to your sample string. For example, in PostgreSQL (adjust syntax slightly for other languages/environments):

SELECT regexp_replace(
    '{some_text_1,some_text_1_more,some_text_1_more_more,0_1_2}INTERSECT{some_text_2,some_text_2_more,some_text_2_more_more,text,abc_efg_hijk}',
    '_(?=[^},]*})(?<=\{.*,)',
    '-', -- Replace with your preferred delimiter here
    'g' -- Global flag to replace all matching underscores
);

Result

After replacement, your string will become:

{some_text_1,some_text_1_more,some_text_1_more_more,0-1-2}INTERSECT{some_text_2,some_text_2_more,some_text_2_more_more,text,abc-efg-hijk}

Handling Edge Cases

If you need to handle {...} blocks that have no commas (i.e., the entire block is a single segment where you want to replace all underscores), adjust the regex to: