Linux Bash：按特殊字符优先排序指定子串的实现方案

我需要对输入文件按第8个“/”后的子串（即“svc...”开头的部分）排序，且该子串内需按以下优先级排序：1. 字符“:”，2. 字符“/”，3. 字母字符。请问如何用sort命令或其他方法实现？

输入文件：

/path1/path2/path3/path4/path5/path6/ser/svcba/log:
/path1/path2/path3/path4/path5/path6/ser/svcba:
/path1/path2/path3/path4/path5/path6/ser/svcbb/log:
/path1/path2/path3/path4/path5/path6/ser/svcbc/log:
/path1/path2/path3/path4/path5/path6/ser/svcbc:
/path1/path2/path3/path4/path5/path6/ser/svcb/log: 
/path1/path2/path3/path4/path5/path6/ser/svcb:
/path1/path2/path3/path4/path5/path6/ser/svca/log: 
/path1/path2/path3/path4/path5/path6/ser/svca:
/path1/path2/path3/path4/path5/path6/ser/svcbb:

期望输出：

/path1/path2/path3/path4/path5/path6/ser/svca:
/path1/path2/path3/path4/path5/path6/ser/svca/log: 
/path1/path2/path3/path4/path5/path6/ser/svcb:
/path1/path2/path3/path4/path5/path6/ser/svcb/log: 
/path1/path2/path3/path4/path5/path6/ser/svcba:
/path1/path2/path3/path4/path5/path6/ser/svcba/log:
/path1/path2/path3/path4/path5/path6/ser/svcbb:
/path1/path2/path3/path4/path5/path6/ser/svcbb/log:
/path1/path2/path3/path4/path5/path6/ser/svcbc:
/path1/path2/path3/path4/path5/path6/ser/svcbc/log:

已尝试但未达预期的命令：

# 直接sort，带/log的行排在不带的前面，不符合需求
cat input.txt | sort
/path1/path2/path3/path4/path5/path6/ser/svca/log: 
/path1/path2/path3/path4/path5/path6/ser/svca:
/path1/path2/path3/path4/path5/path6/ser/svcba/log:
/path1/path2/path3/path4/path5/path6/ser/svcba: 
/path1/path2/path3/path4/path5/path6/ser/svcbb/log:
/path1/path2/path3/path4/path5/path6/ser/svcbb: 
/path1/path2/path3/path4/path5/path6/ser/svcbc/log:
/path1/path2/path3/path4/path5/path6/ser/svcbc: 
/path1/path2/path3/path4/path5/path6/ser/svcb/log: 
/path1/path2/path3/path4/path5/path6/ser/svcb:

# 以:为分隔符排序，svcb/log的位置错误
cat input.txt | sort -t : -k 1,1
/path1/path2/path3/path4/path5/path6/ser/svca:
/path1/path2/path3/path4/path5/path6/ser/svca/log: 
/path1/path2/path3/path4/path5/path6/ser/svcb:
/path1/path2/path3/path4/path5/path6/ser/svcba:
/path1/path2/path3/path4/path5/path6/ser/svcba/log:
/path1/path2/path3/path4/path5/path6/ser/svcbb:
/path1/path2/path3/path4/path5/path6/ser/svcbb/log:
/path1/path2/path3/path4/path5/path6/ser/svcbc:
/path1/path2/path3/path4/path5/path6/ser/svcbc/log:
/path1/path2/path3/path4/path5/path6/ser/svcb/log: 

# 以/为分隔符取第9段排序，带/log的行仍排在前面
cat input.txt | sort -t / -k 9,9  
/path1/path2/path3/path4/path5/path6/ser/svca/log: 
/path1/path2/path3/path4/path5/path6/ser/svca:
/path1/path2/path3/path4/path5/path6/ser/svcb/log: 
/path1/path2/path3/path4/path5/path6/ser/svcba/log:
/path1/path2/path3/path4/path5/path6/ser/svcba: 
/path1/path2/path3/path4/path5/path6/ser/svcbb/log:
/path1/path2/path3/path4/path5/path6/ser/svcbb: 
/path1/path2/path3/path4/path5/path6/ser/svcbc/log:
/path1/path2/path3/path4/path5/path6/ser/svcbc: 
/path1/path2/path3/path4/path5/path6/ser/svcb: 

可以通过awk预处理生成自定义排序键，再结合sort实现需求，具体命令如下：

awk -F '/' '{key=$9; gsub(/:/, "\x01", key); gsub(/\//, "\x02", key); print key "\t" $0}' input.txt | sort | cut -f2-

命令解释： #

1. awk部分：

   • -F '/'：以/作为字段分隔符，第8个/后的内容对应第9个字段$9
   • gsub(/:/, "\x01", key)：把排序键中的:替换成ASCII值为1的控制字符（该字符比/和所有字母的ASCII值都小，排序时会排在最前）
   • gsub(/\//, "\x02", key)：把排序键中的/替换成ASCII值为2的控制字符（比字母小，比\x01大，排序时排在:之后、字母之前）
   • print key "\t" $0：将自定义排序键和原行内容用制表符分隔输出

2. sort：对awk输出的内容按排序键进行字典序排序，此时:对应的\x01会优先于/对应的\x02，最后是字母

3. cut -f2-：去掉前面的自定义排序键，只保留原行内容，得到最终的排序结果

执行该命令后，输出将与你期望的结果完全一致。

内容的提问来源于Stack Exchange，提问作者wasilis