UNIX for Dummies Questions & Answers

Hi,

For an order I requested, the provider has uploaded a tar file in public FTP site which internally has tons of files (compressed) and I need to download files that follows particular pattern which would be few hundreds.

Note: The order can't be requested for files that follows the pattern.

Can this be performed using CURL or WGET commands? Please advise.

Thanks,
Amalan J

Provided the pattern you need is relativly simple (ie file globbing rather than full regex), you can pass wildcards via the FTP protocol.

For a vanilla commandline ftp, you'd want to turn off prompting (tpye 'prompt' until you see it's set to off), while you are at it, turn on hash ('hash') and check that your binary mode is correct for you ('bin' or 'asc' depending on if you want it to convert carriage returns/line feeds). Then use mget to specify your pattern.

mget is nifty in that it will do multiple gets but it can't do filename renaming/relocating on the fly so make sure you are sitting in the right directory ('lcd' before your run it).

curl and wget aren't really right for the task but you could possibly use a tool like that to first get the directory, then grep the resultant index for your files and feed that into a loop... but... yuck :/

That said, I've heard talk in the curl dev lists of adding support for ftp wildcards, so maybe recent versions do this (man curl and see how you go)

Thanks for the reply.
But my need is to get multiple files within a single tar file available in ftp. And I need to know is there a way to extract only those files I need from tar instead of downloading the entire tar file.
Thanks.

Download in ftp is file-based, so you can only download a file or not download it, but you cannot download selected parts of it. You will have to download the tar-file, unpack it locally and then throw away all files you do not need.

I hope this helps.

bakunin

Weeeellll, with a bit of bodging you can sort of do this.
Tar allows you to operate on an incoming stream (it was originally designed for this) and stop when you've got all you want. You still need to seek through the file until you've found all the files you want:

Code:

wget -q -O - "<url of tarball>" | gunzip -c | tar -xf - "<filename you want to extract>"

This will seek through the tarball, streaming off the ftp server and pop the file you want off as it passes. You can specify multiple files on the one commandline but this will only work if you know the files (although you could try passing an escaped wildcard and see how you go - you'll have to manually abort the extract once you feel you have all the files you want though).
You can also pass the "--files-from <file containing the filnames you want>" flag to tar if you prefer.

Note that this won't save you loads of time, especially so if all the files you want are at the end of the archive, but it will help if they are at the start of the file and it's an interesting exercise in streams.

This should help me a lot as it might save 50% of my time and space. Thanks for your help.

Maybe only 50% of your space. I am not sure but "tar" might read the file to the end unless you manually intervent.

Still, this is a very clever idea of Smiling Dragon. I filed it away immediately in my "all-things-interesting-to-know"-notes.

bakunin