在bash脚本中的冒号后查找并替换值

Question

I have a file like this

parameters:
    database_host: 127.0.0.1
    database_port: null
    database_name: myDb
    database_user: root
    database_password: null
    mailer_transport: smtp
    mailer_host: 127.0.0.1
    mailer_user: null
    mailer_password: null
    secret: ea64f518be08e0d5895335990f10d984c22f400c
    name_db: newDB

This is a yml file, and it has a particular format

I Want to create a file sh that find and replace the value after name_db with a parameter

Answer 1

Look for the key ( name_db: ) and replace the value after it. Assuming a sed that supports the -i option:

sed -i.bak "/^[[:space:]]*name_db:/ s/:.*/: $newname/" file.yml

The regular expression /^[[:space:]]*name_db:/ looks for a line that starts with zero or more spaces followed by name_db: . When that line is found, the s/:.*/: $newname/ substitute is executed. It will replace the colon and whatever follows with colon, space, and the value in $newname . This will fail if $newname contains a slash. If that's a possibility, choose a different character (other than / ) as the marker. In case of doubt, Control-A is quite useful and unlikely to be part of a valid name_db value.

This doesn't care what the old value is. You can simply wrap this whole expression in double quotes, which interpolates the value in $newname . Nevertheless, using single quotes around sed scripts is generally a good idea.

---

monkeyUser commented:

sed: 1: "file.yml": extra characters at the end of p command

Note that the command line shown (using -i.bak ) works on both Linux (GNU sed ) and Mac OS X (BSD sed ). GNU sed allows an optional backup suffix which must be attached to the -i option if it is present; BSD sed requires a suffix which must either be attached to the -i option as shown or can be the next argument. If you want no backup with GNU sed , specify just -i … ; with BSD sed , specify '' as an empty argument after the -i option: -i '' … . Given the error message sed: 1: "file.yml": extra characters at the end of p command , I'm suspicious that the code was run on a system with BSD sed , though I'm not quite sure why p was mentioned (I suspect some editing of the actual error message). That is: sed -i /something/ file.yml would, with BSD sed , treat /something/ as the backup file suffix and then find problems with file.yml treated as a sed script…except I think the file name began with p rather than f , since f isn't a sed command (but p is).