只需将其与适当的正则表达式匹配即可。

例如GNU sed（对于posixish sed只需将\+替换为\{1,\}）如下：

sed 'N;s/\([^ ]*[ ]\+\)\{0,1\}\([^ =]\+\)[ ]*=[ ]\{0,1\}\([^ ]\+\) [^ ]* \([^ =]\+\)[ ]*=[ ]*\([^ ]\+\)[^\n]*\n\([^ ]\+\) {\([^}]\+\)}.* [^ ]\+ - [^ ]\+ .* \([^ ]\+\) - \([^ ]\+\).*/\2 = \3,\4 = \5,\6 = \7,\8 = \9/' <<EOF
key1=value1 AND key2=value2 followed by some other text
key3 {value3} some text key4 - value4 and key5 - value5
abc = 12ty3 AND jfk = 345 followed by some other text
klm {678er} some text plr - 567 and deg - 345
xyz-232  abc = 126y3 AND jfk = 567 followed by some other text dre {567x}
klm {rtyyr} some text plr - 444 and deg - 555 some text 345 = uut
EOF

似乎可以正常工作并生成以下输出：

key1 = value1,key2 = value2,key3 = value3,key5 = value5
abc = 12ty3,jfk = 345,klm = 678er,deg = 345
abc = 126y3,jfk = 567,klm = rtyyr,deg = 555

我将使用grep -o提取每个事件在自己的行上，sed将需要的内容重新格式化，并使用paste将它们重新合并为一行：

grep -Eo '\w+\s*=\s*\w+|\w+\s+\{[^}]+\}|\w+\s+-\s+\w+' | sed -E 's/-/=/;s/\{([^}]+)}/= \1/' | paste -sd ','

您可以try it here。