search and replace in whole directory recursively

Question

I have many entities in my java project like this:

public class MyCar {
// The brand of the car
private String brand;
// The time owner bought it
private Date obtainTime;
// The price in dollar
private Double price;
}

Now I want to replace them with this format:

public class MyCar {
@ApiModelProperty("The brand of the car")
private String brand;
@ApiModelProperty("The time owner bought it")
private Date obtainTime;
@ApiModelProperty("The price in dollar")
private Double price;
}

Is there any way to do this in my whole project, which means to search and replace in whole directory recursively?

Answer 1

You have a couple of good answers here already, but the fact that both solutions use the shell implies that you can't do this from inside Vim, which isn't true! Here are two possible Vim-native solutions:

Use vimgrep

1. Add all the lines to change to your quickfix:

    :vimgrep /\v\/\/\s*.*\n(private|public)(\s+\k+){2};/ **/*.java
   

2. For each line, make the edit with normal mode commands:

    :cdo norm!caw@ApiModelProperty("^[A")
                                    ^^
                                    This is a literal ESC character.
                                    Enter it by typing <Ctrl-V><Esc>
   

You could of course use a :substitute) command to make the edit in step 2. I just didn't fancy writing out the regular expression three times.

Use the arglist

1. Set your arglist to contain all the .java files:

    :args **/*.java
   

2. Make a substitution for every match in every file:

    :argdo %s/\v\/\/\s*(.*)\ze\n(private|public)(\s+\k+){2};/@ApiModelProperty("\1")/ | up
   

If you have the 'hidden' option set, you don't need the | up at the end of the second command. This is only required to write each file if there are changes before switching to the next file, otherwise Vim will complain.

Answer 2

I have two shell functions in my ~/.zshrc that are quite useful for this (I think they should work for bash as well, but I didn't test, and with minimal modification it should even work in standard POSIX shells):

# "ag edit" and "grep edit".
age() {
    IFS=$'\n' files=($(ag "$@" | cut -d: -f1 | sort -u))
    [[ -z "$files" ]] && return 1
    vim \
        +':silent! /\v'"${@[-1]/\//\\/}" \
        +':silent! tabdo :1 | normal! n' \
        +':tabfirst' \
        -p $files
}
grepe() {
    IFS=$'\n' files=($(ag "$@" | cut -d: -f1 | sort -u))
    [[ -z "$files" ]] && return 1
    vim \
        +':silent! /\v'"${@[-1]/\//\\/}" \
        +':silent! tabdo :1 | normal! n' \
        +':tabfirst' \
        -p $files
}

What I usually do is ag 'String field' first to confirm that it matches what I expect, and then modify this to age 'String field' to open all the files in Vim tabs (vim -p opens every file in a new tab).

This sets the search pattern in Vim to whatever you typed on the commandline and goes to the first match, but this isn't perfect as the Vim syntax isn't the same as the syntax grep and ag use, so complex patterns tend to break. It's usually "good enough" for simpler patterns.

Then :%s/.../ whatever you want, and if you're confident you've got it right run :tabdo :%s/.../ so that it's run on all tabs. :wqa and verify that it's all correct.

You can use ripgrep (rg) instead of the_silver_searcher (ag) as well; actually, ag an alias for rg for me, it's just that I was already so used to typing ag when I switched from ag to rg that it made sense to keep using this command.

You don't need to use tabs, you can use buffers and :bufdo as well. I just like to use tabs.

There are a myriad of other ways to do the same, B Layer's answer already mentioned some. But I find that my custom age is super helpful, for this use case and many others, and probably the "custom command" I use the most.

Answer 3

filbranden asks some good questions but since I've already got this ready we can use it as a jump off point. Hopefully nothing more than tweaking of the substitution command will be needed.

I'd recommend that you not try to do this from within an instance of Vim but instead leverage the power of a shell (Bash) PLUS Vim. (If you're on Windows then this will work with Cygwin or WSL. If you don't have either, then, well...we're out of luck.)

This is one way to do it...

while IFS= read -r f; do echo "$f"; vim -e -c '%s#\v^// (field comment)#@ApiModelProperty("\1")#g' -c 'x' "$f"; done < <(find {DIR} -name \*.java)

For easier reading here's it broken out into multiple lines...

while IFS= read -r f; do
    echo "Processing $f ..."
    vim -e -c '%s#\v^// (field comment)#@ApiModelProperty("\1")#g' -c 'x' "$f"
done < <(find {DIR} -name \*.java)

Change {DIR} to the path of the root directory of your project. Or cd to that dir before running the command and just use . in place of {DIR}.

The while loop is in a form that handles funky file names pretty well. It iterates over the output of find. The find command will recursively search for files with extension java. Each of these will be passed to an invocation of Vim in "batch mode" (that's what the -e is for). The -c flag tells Vim to run what follows as a regular Ex command and I'm doing that with a standard substitution and again with x which writes the file, if necessary, and quits editing the current file.

Note: A lot of people might opt to use sed -i here instead of Vim but actually Vim is better suited for this task since it modifies the file without having to clumsily make a copy of the target file like sed does.

BTW, if you really wanted to you could call this from within Vim but you might have to make some changes to the quoting or add some escaping.