SftpSession.exists() does not throw exception

[gary-x-li]

SftpSession.exists() method catches SftpException but ignores it. Why not throw a NestedIOException like other methods do?

In my code I use exists to check whether the remote file exists or not before I call remove to delete the file. In case when the logic inside exists throws an IOException (Pipe closed), it returns false, indicating the remote file does not exists.

Thanks,
Gary

[artembilan]

We need to think why is that exception ignored so far because the session contract is:

	/**
	 * Check if the remote file or directory exists.
	 * @param path the remote path.
	 * @return {@code true} or {@code false} if remote path exists or not.
	 * @throws IOException an IO exception during remote interaction.
	 */
	boolean exists(String path) throws IOException;

Meanwhile you can consider a workaround for your use-case like:

SftpRemoteFileTemplate remoteFileTemplate = new SftpRemoteFileTemplate(sessionFactory());
remoteFileTemplate.<Boolean, ChannelSftp>executeWithClient(client -> {
				client.lstat(path);
				return true;
		});

So, if an exception is thrown from that lstat() call, it won't return false for you, but will be re-thrown to your code.

[artembilan]

I think our solution must be like this: https://stackoverflow.com/questions/13202789/how-to-check-directory-is-existed-before-creating-a-new-directory-in-jsch

Where a logical answer is present in the comment from Ton Bosma:

I think Exception is to eager, it can be more specific:

try{
      channel.stat(folderName); 
   } 
   catch (SftpException e ) { 
       if (e.id == ChannelSftp.SSH_FX_NO_SUCH_FILE ) {
             channel.mkdir(folderName);
       }
       else {
         throw e;
       } 
  }

Does it make sense?

WDYT, @garyrussell ?

Thanks

[artembilan]

Even if I treat this as a enhancement I think we need to back-port it to correct a behavior according contract expectations.

[garyrussell]

I assume you mean...

if (e.id == ChannelSftp.SSH_FX_NO_SUCH_FILE ) {
    return false;
}
else {
    throw e;
}

Yes, I agree it should be back-ported even though it's a behavior change.

[artembilan]

Yes, sorry, copy/paste artifact.

PR is on its way. Stay tuned...