使用`find`查找pom.xml文件并进行Maven构建，忽略嵌套的POM

Question

I currently have a couple of applications with different folder structures. I am using maven for building these apps. I am running into some issues. At times the main pom.xml file is in the root directory and other times it is in the sub-directory. I only want to do a maven build with the main pom.xml found in the main/highest directory. I am running the find command from workspace (below for more details). How could I achieve this without depending on the folder structure?

Shell Command:

find * -maxdepth 1 -name pom.xml -execdir mvn clean package -Dmaven.test.error.ignore=true -Dmaven.test.failure.ignore=true \;

This works with this folder structure:

workspace
|-- app1
|   |-- pom.xml
|   |-- abrt.conf
|   |-- gpg_keys
|   `-- web
|       |-- pom.xml
|       `-- test.conf

returns:

app1/pom.xml

The command does returns a couple of pom.xml files with this folder structure:

workspace
|   |-- pom.xml
|   |-- abrt.conf
|   |-- gpg_keys
|   `-- web
|       |-- pom.xml
|       `-- test.conf

returns:

web/pom.xml
pom.xml

Answer 1

Because find operates breadth-first unless explicitly told to do otherwise, the first result will always be the shallowest. Thus, you just need to tell it to quit after finding a single POM file.

In GNU find , there's a -quit action for exactly this purpose:

find . -name pom.xml \
  -execdir mvn clean package -Dmaven.test.error.ignore=true -Dmaven.test.failure.ignore=true \; \
  -quit

If you aren't guaranteed to have GNU find , but are guaranteed to have bash, you can read just the first result and operate on that:

if IFS= read -r -d '' pom_filename < <(find . -name pom.xml -print0); then
  (cd -- "${pom_filename%/*}" && exec mvn clean package -Dmaven.test.error.ignore=true -Dmaven.test.failure.ignore=true)
fi

Note the use of -print0 -- that way we avoid maliciously-named directories (with newlines) being able to influence behavior by injecting extra names into the list.

---

Now, let's consider a trickier case:

A/pom.xml
A/sub/pom.xml
B/pom.xml
B/sub/pom.xml

If you want to run for both A and B, but not A/sub or B/sub, then things get a bit more interesting:

find . \
    -type d -exec test -e '{}/pom.xml' ';' \
    -execdir mvn clean pkg ... ';' \
    -prune

Note that we're using -prune to tell find to stop recursing into any directory where a pom.xml exists, but only after we already ran a build there.

Answer 2

Just do a find without the * (replace with . as suggested in comment) to list from highest directory to the lowest. Pipe to head -1 to get the first one of the list. Put all of this into a command substitution ( $(...) ) to assign a variable. Run the same command with your variable instead. Since you may have / in the file name now, we will not need -name , -maxdepth or * .

pomxmlfile="$(find . -name pom.xml | head -1)"
find "$pomxmlfile" -execdir mvn clean package -Dmaven.test.error.ignore=true -Dmaven.test.failure.ignore=true \;

Answer 3

This line finds the pom.xml with the shortest path (ie the pom nearest the top of the directory tree) and executes mvn from its directory:

find . -name pom.xml | awk '{print length($0) " " $0}' | sort -n | cut -d ' ' -f 2- | head -n 1 | xargs -I {} bash -c 'cd $(dirname {}) && mvn clean package -Dmaven.test.error.ignore=true -Dmaven.test.failure.ignore=true \;'

Pipe chain breakdown:

• finds all pom.xml under current directory
• adds length of line to start of line
• sorts lines by number (ie the length) lowest first
• removes the length
• keeps only the first (ie shortest) line
• runs mvn from the directory of the pom.xml file

Answer 4

I'm not sure. If the main pom.xml is in a directory a/b/c and no other pom.xml is in a subdirectory below a other than a/b/c , then the main pom.xml must have the shortest path string. So in that case I guess this command brings up the pom.xml closest to root and can be executed from either a , a/b , or a/b/c :

find -name pom.xml | awk 'NR==1 || length() < length(s) { s = $0 } END { print s }'