Bash: how to automatically insert spacing characters according to indent?

I've got lots of debug statements making for unreadable stacktraces (not my call) like this:

 00:53:59,906  - j.util.indexing.FileBasedIndex - START INDEX SHUTDOWN 
 00:53:09,192  - .impl.stores.XmlElementStorage - Document was not loaded for $APP_CONFIG$/macros.xml 
 00:53:09,195  - s.impl.stores.FileBasedStorage - Document was not loaded for $APP_CONFIG$/quicklists.xml file is null 
 00:53:09,195  - .impl.stores.XmlElementStorage - Document was not loaded for $APP_CONFIG$/quicklists.xml 
 00:53:09,696  - ij.openapi.wm.impl.IdeRootPane - App initialization took 5584 ms 
 00:53:11,677  -                  TestNG Runner - Create TestNG Template Configuration 
 00:53:13,628  - indexing.UnindexedFilesUpdater - Unindexed files update started 
 00:53:15,370  - indexing.UnindexedFilesUpdater - Unindexed files update done 
 00:53:20,873  - tor.impl.FileEditorManagerImpl - Project opening took 10569 ms 
 00:53:31,862  - s.impl.stores.FileBasedStorage - Document was not loaded for $APP_CONFIG$/intentionSettings.xml file is null 
 00:53:31,862  - .impl.stores.XmlElementStorage - Document was not loaded for $APP_CONFIG$/intentionSettings.xml 
 00:54:00,723  - j.util.indexing.FileBasedIndex - END INDEX SHUTDOWN

And I want them to look like this:

 00:53:59,906  - j.util.indexing.FileBasedIndex - START INDEX SHUTDOWN 
 00:53:09,192  - .impl.stores.XmlElementStorage -   Document was not loaded for $APP_CONFIG$/macros.xml 
 00:53:09,195  - s.impl.stores.FileBasedStorage -   Document was not loaded for $APP_CONFIG$/quicklists.xml file is null 
 00:53:09,195  - .impl.stores.XmlElementStorage -   Document was not loaded for $APP_CONFIG$/quicklists.xml 
 00:53:09,696  - ij.openapi.wm.impl.IdeRootPane -     App initialization took 5584 ms 
 00:53:11,677  -                  TestNG Runner -       Create TestNG Template Configuration 
 00:53:13,628  - indexing.UnindexedFilesUpdater -        Unindexed files update started 
 00:53:15,370  - indexing.UnindexedFilesUpdater -        Unindexed files update done 
 00:53:20,873  - tor.impl.FileEditorManagerImpl -     Project opening took 10569 ms 
 00:53:31,862  - s.impl.stores.FileBasedStorage -    Document was not loaded for $APP_CONFIG$/intentionSettings.xml file is null 
 00:53:31,862  - .impl.stores.XmlElementStorage -   Document was not loaded for $APP_CONFIG$/intentionSettings.xml 
 00:54:00,723  - j.util.indexing.FileBasedIndex - END INDEX SHUTDOWN

Now my code is formatted automatically so I know exactly how many characters there are in my source code for indent level.

What I want to do is replace automatically every line in my code looking like:

log.debug("some silly debug line");

with:

log.debug(" " + "some silly debug line");

or:

log.debug("  " + "some silly debug line");

depending on the number of indents before the log.debug line.

I'm pretty sure some Bash/shell magic can do this automatically.

Basically I'd start with a:

find . -iname "*java" -exec ...

then I'd like to insert " " on lines starting with log.debug depending on the indent level of these lines.

How would I go about it?

EDIT

Damn: it needs to be able to run several times, without re-shifting the lines already containing log.debug(" " +). It gets more complicated but now thanks to the good answer开发者_StackOverflow社区 I've got something to start from :)

Note: before yelling and shooting as to why this ain't a good idea, read my comment(s)

This should work when run multiple times on files:

sed -i.bak 's/^\([[:blank:]]*\)\(log.debug(\)\("[[:blank:]]*" + \)\?/\1\2"\1" + /; s/^log.debug("" + /log.debug(/'

Combined with your find command:

find . -iname "*java" -exec sed -i.bak 's/^\([[:blank:]]*\)\(log.debug(\)\("[[:blank:]]*" + \)\?/\1\2"\1" + /; s/^log.debug("" + /log.debug(/' \;

Test run:

$ sed -i.bak '...' test.java
$ diff test.java test.java.bak
(report of differences)
$ sed -i.bak '...' test.java
$ diff test.java test.java.bak
(no differences)

Explanation of sed command:

s/^\([[:blank:]]*\)\(log.debug(\)\("[[:blank:]]*" + \)\?/\1\2"\1" + /

• Capture the tabs and spaces, if any, at the beginning of the line into group \1.
• Capture the string "log.debug(" (only so we don't have to repeat it verbatim) into group \2.
• Capture any existing indent string ("indentlevel" +) into so it can be overwritten in case the indentation has changed. The \? makes its existence optional. We don't care that it would be capture group \3 since we're throwing it away. The parentheses are necessary only for the \?.
• Now the replacement is output. Capture groups \1 (the indent) and \2 (the "log.debug(") are put back like they were. And "indentlevel" + is added (where "indentlevel" is the same sequence of whitespace characters that indented the line). Note that this may result in nothing between the quotes.

The next part of the command removes the indentation stuff if it's null. You can omit that if you don't mind resulting lines of the form log.debug("" + "foo") when the line isn't indented.

s/^log.debug("" + /log.debug(/

Try sed string replacement:

find . -type f -iname "*.java" -exec sed -i 's/\(\(\s\+\)log.debug(\)\(.*)\)/\1"\2" + \3/' {} \;

Let's take a closer look at the regex patterns in sed -i 's/\(\(\s\+\)log.debug(\)\(.*)\)/\1"\2" + \3/...

Group 1 is the original leading text, including whitespace and "log.debug(" (with the open parenthesis).
\(\(\s\+\)log.debug(\)

Group 2, which is inside Group 1 shown above, captures only the leading whitespace:
\(\s\+\)

Group 3 is the original remaining text after "log.debug(":
\(.*)\)

sed uses these capture groups to form a replacement string, which begins with the original leading text, following the original leading whitespace surrounded by quotes, then a plus symbol (for Java string concatenation), and finally the remaining original text:
\1"\2" + \3

You can try this awk script:

x.awk

/log.debug/ {
    logStart = match($0, "log");
    spaces = substr($0, 0, logStart);
    gap = "\"" spaces "\"";
    parenStart = match($0, "(");
    firstHalf = substr($0, 0, parenStart);
    lastHalf = substr($0, parenStart, length($0) - parenStart);
    print firstHalf gap " + " lastHalf;
    next;
}
{
    print;
}

awk -f x.awk orig.java > new.java

Easy way to do this would be:

cd /code/root
mkdir /tmp/newcode
cp -R * /tmp/newcode
for i in `find . -name "*.java"`
do
    awk -f x.awk > /tmp/newcode/$i
done

Should just copy over the original code on top of the new code, and not have to fight with making directories or any of that.

Untested, there can easily be a one off error in the splitting of the line, but it should get your started.

And you were right!¹

sed -i .tmp -e 's/^\( *\)log.debug(/\1log.debug("\1" + /' prog.java

Integrated with find(1) it becomes:

find . -name \*.java -a -exec sed -i .tmp -e 's/^\( *\)log.debug(/\1log.debug("\1" + /' {} \;

This finds everything named *.java and runs an in-place sed (stream editor) program, matches the lines to change with a \( capture group \) around the space string, and then it tacks the capture group back in where it found it as well as in front of your string.

The -a, -exec, \; and {} are just the arcane find syntax for "and", "execute program", "end the command being executed" and "substitute the found filename here".

^{1. When you said: "I'm pretty sure some Bash/shell magic can do this automatically."}