pwget(1) Perl Web URL fetch program


pwget [URL ...]
pwget --config $HOME/config/pwget.conf --Tag linux --Tag emacs ..
pwget --verbose --overwrite
pwget --verbose --overwrite --Output ~/dir/
pwget --new --overwrite


Automate periodic downloads of files and packages.

If you retrieve latest versions of certain program blocks periodically, this is the Perl script for you. Run from cron job or once a week to upload newest versions of files around the net. Note:

Wget and this program

At this point you may wonder, where would you need this perl program when wget(1) C-program has been the standard for ages. Well, 1) Perl is cross platform and more easily extendable 2) You can record file download criterias to a configuration file and use perl regular epxressions to select downloads 3) the program can anlyze web-pages and ``search'' for the download only links as instructed 4) last but not least, it can track newest packages whose name has changed since last downlaod. There are heuristics to determine the newest file or package according to file name skeleton defined in configuration.

This program does not replace pwget(1) because it does not offer as many options as wget, like recursive downloads and date comparing. Use wget for ad hoc downloads and this utility for files that change (new releases of archives) or which you monitor periodically.

Short introduction

This small utility makes it possible to keep a list of URLs in a configuration file and periodically retrieve those pages or files with simple commands. This utility is best suited for small batch jobs to download e.g. most recent versions of software files. If you use an URL that is already on disk, be sure to supply option --overwrite to allow overwriting existing files.

While you can run this program from command line to retrieve individual files, program has been designed to use separate configuration file via --config option. In the configuration file you can control the downloading with separate directives like "save:" which tells to save the file under different name. The simplest way to retrieve the latest version of apackage from a FTP site is:

    pwget --new --overwite --verbose \

Do not worry about the filename "package-1.00.tar.gz". The latest version, say, "package-3.08.tar.gz" will be retrieved. The option --new instructs to find newer version than the provided URL.

If the URL ends to slash, then directory list at the remote machine is stored to file:


The content of this file can be either index.html or the directory listing depending on the used http or ftp protocol.


-A, --regexp-content REGEXP
Analyze the content of the file and match REGEXP. Only if the regexp matches the file content, then download file. This option will make downloads slow, because the file is read into memory as a single line and then a match is searched against the content.

For example to download Emacs lisp file (.el) written by Mr. Foo in case insensitive manner:

    pwget -v -R '\.el$' -A "(?i)Author: Mr. Foo" \

-C, --create-paths
Create paths that do not exist in "lcd:" directives.

By default, any LCD directive to non-existing directory will interrupt program. With this option, local directories are created as needed making it possible to re-create the exact structure as it is in configuration file.

-c, --config FILE
This option can be given multiple times. All configurations are read.

Read URLs from configuration file. If no configuration file is given, file pointed by environment variable is read. See ENVIRONMENT.

The configuration file layout is envlained in section CONFIGURATION FILE

Do a chdir() to DIRECTORY before any URL download starts. This is like doing:


-d, --debug [LEVEL]
Turn on debug with positive LEVEL number. Zero means no debug. This option turns on --verbose too.
-e, --extract
Unpack any files after retrieving them. The command to unpack typical archive files are defined in a program. Make sure these programs are along path. Win32 users are encouraged to install the Cygwin utilities where these programs come standard. Refer to section SEE ALSO.

  .tar => tar
  .tgz => tar + gzip
  .gz  => gzip
  .bz2 => bzip2
  .zip => unzip
-F, --firewall FIREWALL
Use FIREWALL when accessing files via ftp:// protocol.
-h, --help
Print help page in text.
Print help page in HTML.
Print help page in Unix manual page format. You want to feed this output to c<nroff -man> in order to read it.

Print help page.

-m, --mirror SITE
If URL points to Sourcefoge download area, use mirror SITE for downloading. Alternatively the full full URL can include the mirror information. And example:

    --mirror kent

-n, --new
Get newest file. This applies to datafiles, which do not have extension .asp or .html. When new releases are announced, the version number in filename usually tells which is the current one so getting harcoded file with:

    pwget -o -v

is not usually practical from automation point of view. Adding --new option to the command line causes double pass: a) the whole is examined for all files and b) files matching approximately filename program-1.3.tar.gz are examined, heuristically sorted and file with latest version number is retrieved.

Ignore "lcd:" directives in configuration file.

In the configuration file, any "lcd:" directives are obeyed as they are seen. But if you do want to retrieve URL to your current directory, be sure to supply this option. Otherwise the file will end to the directory pointer by "lcd:".

Ignore "save:" directives in configuration file. If the URLs have "save:" options, they are ignored during fetch. You usually want to combine --no-lcd with --no-save
Ignore "x:" directives in configuration file.
-O, --output DIR
Before retrieving any files, chdir to DIR.
-o, --overwrite
Allow overwriting existing files when retrieving URLs. Combine this with --skip-version if you periodically update files.
--proxy PROXY
Use PROXY server for HTTP. (See --Firewall for FTP.). The port number is optional in the call:

-p, --prefix PREFIX
Add PREFIX to all retrieved files.
-P, --postfix POSTFIX
Add POSTFIX to all retrieved files.
-D, --prefix-date
Add iso8601 ``:YYYY-MM-DD'' prefix to all retrived files. This is added before possible --prefix-www or --prefix.
-W, --prefix-www
Usually the files are stored with the same name as in the URL dir, but if you retrieve files that have identical names you can store each page separately so that the file name is prefixed by the site name.    -->   -->
-r, --regexp REGEXP
Retrieve file matching at the destination URL site. This is like ``Connect to the URL and get all files matching REGEXP''. Here all gzip compressed files are found form HTTP server directory:

    pwget -v -R "\.gz"

-R, --config-regexp REGEXP
Retrieve URLs matching REGEXP from configuration file. This cancels --Tag options in the command line.
-s, --selftest
Run some internal tests. For maintainer or developer only.
--sleep SECONDS
Sleep SECONDS before next URL request. When using regexp based downlaods that may return many hits, some sites disallow successive requests in within short period of time. This options makes program sleep for number of SECONDS between retrievals to overcome 'Service unavailable'.
Retrieve URL and write to stdout.
Do not download files that have version number and which already exists on disk. Suppose you have these files and you use option --skip-version:


Only file.txt is retrieved, because file-1.1.tar.gz contains version number and the file has not changed since last retrieval. The idea is, that in every release the number in in distribution increases, but there may be distributions which do not contain version number. In regular intervals you may want to load those packages again, but skip versioned files. In short: This option does not make much sense without additional option --new

If you want to reload versioned file again, add option --overwrite.

-t, --test, --dry-run
Run in test mode.
-T, --tag NAME [NAME] ...
Search tag NAME from the config file and download only entries defined under that tag. Refer to --config FILE option description. You can give Multiple --Tag switches. Combining this option with --regexp does not make sense and the concequencies are undefined.
-v, --verbose [NUMBER]
Print verbose messages.
-V, --version
Print version information.


Get files from site:

    pwget ..

Display copyright file for package GNU make from Debian pages:

    pwget --stdout --regexp 'copyright$'

Get all mailing list archive files that match ``gz'':

    pwget --regexp gz

Read a directory and store it to filename YYYY-MM-DD::!dir!000root-file.

    pwget --prefix-date --overwrite --verbose

To update newest version of the package, but only if there is none at disk already. The --new option instructs to find newer packages and the filename is only used as a skeleton for files to look for:

    pwget --overwrite --skip-version --new --verbose \

To overwrite file and add a date prefix to the file name:

    pwget --prefix-date --overwrite --verbose \

To add date and WWW site prefix to the filenames:

    pwget --prefix-date --prefix-www --overwrite --verbose \

Get all updated files under cnfiguration file's tag updates:

    pwget --verbose --overwrite --skip-version --new --Tag updates
    pwget -v -o -s -n -T updates

Get files as they read in the configuration file to the current directory, ignoring any "lcd:" and "save:" directives:

    pwget --config $HOME/config/pwget.conf /
        --no-lcd --no-save --overwrite --verbose \

To check configuration file, run the program with non-matching regexp and it parses the file and checks the "lcd:" directives on the way:

    pwget -v -r dummy-regexp
    pwget.DirectiveLcd: LCD [$EUSR/directory ...]
    is not a directory at /users/foo/bin/pwget line 889.



The configuration file is NOT Perl code. Comments start with hash character (#).


At this point, variable expansions happen only in lcd:. Do not try to use them anywhere else, like in URLs.

Path variables for lcd: are defined using following notation, spaces are not allowed in VALUE part (no directory names with spaces). Varaible names are case sensitive. Variables substitute environment variabales with the same name. Environment variables are immediately available.

    VARIABLE = /home/my/dir         # define variable
    VARIABLE = $dir/some/file       # Use previously defined variable
    FTP      = $HOME/ftp            # Use environment variable

The right hand can refer to previously defined variables or existing environment variables. Repeat, this is not Perl code although it may look like one, but just an allowed syntax in the configuration file. Notice that there is dollar to the right hand> when variable is referred, but no dollar to the left hand side when variable is defined. Here is example of a possible configuration file contant. The tags are hierarchically ordered without a limit.

Warning: remember to use different variables names in separate include files. All variables are global.

Include files

It is possible to include more configuration files with statement

    INCLUDE <path-to-file-name>

Variable expansions are possible in the file name. There is no limit how many or how deep include structure is used. Every file is included only once, so it is safe to to have multiple includes to the same file. Every include is read, so put the most importat override includes last:

    INCLUDE <etc/pwget.conf>             # Global
    INCLUDE <$HOME/config/pwget.conf>    # HOME overrides it

A special "THIS" tag means relative path of the current include file, which makes it possible to include several files form the same directory where a initial include file resides

    # Start of config at /etc/pwget.conf
    # THIS = /etc, current location
    include <THIS/pwget-others.conf>
    # Refers to directory where current user is: the pwd
    include <pwget-others.conf>
    # end

Configuraton file example

The configuration file can contain many <directoves:>, where each directive end to a colon. The usage of each directory is best explained by examining the configuration file below and reading the commentary near each directive.

    #   $HOME/config/pwget.conf F- Perl pwget configuration file
    ROOT   = $HOME                      # define variables
    CONF   = $HOME/config
    UPDATE = $ROOT/updates
    DOWNL  = $ROOT/download
    #   Include more configuration files. It is possible to
    #   split a huge file in pieces and have "linux",
    #   "win32", "debian", "emacs" configurations in separate
    #   and manageable files.
    INCLUDE <$CONF/pwget-other.conf>
    INCLUDE <$CONF/pwget-more.conf>
    tag1: local-copies tag1: local      # multiple names to this category
        lcd:  $UPDATE                   # chdir directive
        #  This is show to user with option --verbose
        print: Notice, this site moved YYYY-MM-DD, update your bookmarks
    tag1: external
      lcd:  $DOWNL
      tag2: external-http save:/dir/dir/page.html
      tag2: external-ftp save:xx-file.txt.gz login:foo pass:passwd x:
        lcd: $HOME/download/package new:
      tag2: package-x
        lcd: $DOWNL/package-x
        #  Person announces new files in his homepage, download all
        #  announced files. Unpack everything (x:) and remove any
        #  existing directories (xopt:rm) pregexp:\.tar\.gz$ x: xopt:rm
    # End of configuration file pwget.conf


All the directives must in the same line where the URL is. The programs scans lines and determines all options given in line for the URL. Directives can be overridden by command line options.
Currently only conv:text is available.

Convert downloaded page to text. This option always needs either save: or rename:, because only those directives change filename. Here is an example: cnv:text save:file.txt pregexp:\.html cnv:text rename:s/html/txt/

A text: shorthand directive can be used instead of cnv:text.

Download file only if the content matches REGEXP. This is same as option --Regexp-content. In this example directory listing Emacs lisp packages (.el) are downloaded but only if their content indicates that the Author is Mr. Foo: cregexp:(?i)author:.*Foo pregexp:\.el$
Set local download directory to DIRECTORY (chdir to it). Any environment variables are substituted in path name. If this tag is found, it replaces setting of --Output. If path is not a directory, terminate with error. See also --Create-paths and --no-lcd.
Ftp login name. Default value is ``anonymous''.
This is relevant to Sourceforge only which does not allow direct downloads with links. Visit project's Sourceforge homepage and see which mirrors are available for downloading.

An example: new: mirror:kent
Get newest file. This variable is reset to the value of --new after the line has been processed. Newest means, that an "ls" command is run in the ftp, and something equivalent in HTTP ``ftp directories'', and any files that resemble the filename is examined, sorted and heurestically determined according to version number of file which one is the latest. For example files that have version information in YYYYMMDD format will most likely to be retrieved right.

Time stamps of the files are not checked.

The only requirement is that filename "must" follow the universal version numbering standard:

    FILE-VERSION.extension      # de facto VERSION is defined as [\d.]+
    file-19990101.tar.gz        # ok
    file-1999.0101.tar.gz       # ok
    file-         # ok
    file1234.txt                # not recognized. Must have "-"
    file-0.23d.tar.gz           # warning, letters are problematic

Files that have some alphabetic version indicator at the end of VERSION may not be handled correctly. Contact the developer and inform him about the de facto standard so that files can be retrieved more intelligently.

NOTE: In order the new: directive to know what kind of files to look for, it needs a file tamplate. You can use a direct link to some filename. Here the location ``'' is examined and the filename template used is took as ``file-1.1.tar.gz'' to search for files that might be newer, like ``file-9.1.10.tar.gz'': new:

If the filename appeard in a named page, use directive file: for template. In this case the ``download.html'' page is examined for files looking like ``file.*tar.gz'' and the latest is searched: file:file-1.1.tar.gz new:
overwrite: o:
Same as turning on --overwrite
Read web page and apply commands to it. An example: contact the root page and save it: page: save:foo-homepage.html

In order to find the correct information from the page, other directives are usually supplied to guide the searching.

1) Adding directive "pregexp:ARCHIVE-REGEXP" matches the A HREF links in the page.

2) Adding directive new: instructs to find newer VERSIONS of the file.

3) Adding directive "file:DOWNLOAD-FILE" tells what template to use to construct the downloadable file name. This is needed for the "new:" directive.

4) A directive "vregexp:VERSION-REGEXP" matches the exact location in the page from where the version information is extracted. The default regexp looks for line that says ``The latest version ... is ... N.N''. The regexp must return submatch 2 for the version number.


Search for newer files from a HTTP directory listing. Examine page for model "package-1.1.tar.gz" and find a newer file. E.g. "package-4.7.tar.gz" would be downloaded. new:


Search for newer files from the content of the page. The directive file: acts as a model for filenames to pay attention to. new: pregexp:tar.gz file:package-1.1.tar.gz


Use directive rename: to change the filename before soring it on disk. Here, the version number is attached to the actila filename:


The directived needed would be as follows; entries have been broken to separate lines for legibility:

This effectively reads: ``See if there is new version of something that looks like file.el-1.1 and save it under name file.el by deleting the extra version number at the end of original filename''.


Contact absolute page: at and search A HREF urls in the page that match pregexp:. In addition, do another scan and search the version number in the page from thw position that match vregexp: (submatch 2).

After all the pieces have been found, use template file: to make the retrievable file using the version number found from vregexp:. The actual download location is combination of page: and A HREF pregexp: location.

The directived needed would be as follows; entries have been broken to separate lines for legibility:
    pregexp: package.tar.gz
    vregexp: ((?i)latest.*?version.*?\b([\d][\d.]+).*)
    file: package-1.3.tar.gz

An example of web page where the above would apply:

    The latest version of package is <B>2.4.1</B> It can be
    downloaded in several forms:
        <A HREF="download/files/package.tar.gz">Tar file</A>
        <A HREF="download/files/">ZIP file

For this example, assume that "package.tar.gz" is a symbolic link pointing to the latest release file "package-2.4.1.tar.gz". Thus the actual download location would have been "".

Why not simply download "package.tar.gz"? Because then the program can't decide if the version at the page is newer than one stored on disk from the previous download. With version numbers in the file names, the comparison is possible.

FIXME: This opton is obsolete. do not use.

THIS IS FOR HTTP only. Use Use directive regexp: for FTP protocls.

This is a more general instruction than the page: and vregexp: explained above.

Instruct to download every URL on HTML page matching pregexp:RE. In typical situation the page maintainer lists his software in the development page. This example would download every tar.gz file in the page. Note, that the REGEXP is matched against the A HREF link content, not the actual text that is displayed on the page: page:find pregexp:\.tar.gz$

You can also use additional regexp-no: directive if you want to exclude files after the pregexp: has matched a link. page:find pregexp:\.tar.gz$ regexp-no:desktop
For FTP logins. Default value is "[email protected]".
Search A HREF links in page matching a regular expression. The regular expression must be a single word with no whitespace. This is incorrect:

    pregexp:(this regexp )

It must be written as:

Print associated message to user requesting matching tag name. This directive must in separate line inside tag.

    tag1: linux
      print: this download site moved 2002-02-02, check your bookmarks. new:

The "print:" directive for tag is shown only if user turns on --verbose mode:

    pwget -v -T linux
Rename each file using PERL-CODE. The PERL-CODE must be full perl program with no spaces anywhere. Following variables are available during the eval() of code:

    $ARG = current file name
    $url = complete url for the file
    The code must return $ARG which is used for file name

For example, if page contains links to .html files that are in fact text files, following statement would change the file extensions: page:find pregexp:\.html rename:s/html/txt/

You can also call function "MonthToNumber($string)" if the filename contains written month name, like <2005-February.mbox>.The function will convert the name into number. Many mailing list archives can be donwloaded cleanly this way.

    #  This will download SA-Exim Mailing list archives: pregexp:\.txt$ rename:$ARG=MonthToNumber($ARG)

Here is a more complicated example: pregexp:mbox.*\d$ rename:my($y,$m)=($url=~/year=(\d+).*month=(\d+)/);$ARG="$y-$m.mbox"

Let's break that one apart. You may spend some time with this example since the possiblilities are limitless.

    1. Connect to page
    2. Search page for URLs matching regexp 'mbox.*\d$'. A
       found link could match hrefs like this:
    3. The found link is put to $ARG (same as $_), which can be used
       to extract suitable mailbox name with a perl code that is
       evaluated. The resulting name must apear in $ARG. Thus the code
       effectively extract two items from the link to form a mailbox
        my ($y, $m) = ( $url =~ /year=(\d+).*month=(\d+)/ )
        $ARG = "$y-$m.mbox"
        => 2004-12.mbox

Just remember, that the perl code that follows "rename:" directive must must not contain any spaces. It all must be readable as one string.

Get all files in ftp directory matching regexp. Directive save: is ignored.
After the "regexp:" directive has matched, exclude files that match directive regexp-no:
This option is for interactive use. Retrieve all files from HTTP or FTP site which match REGEXP.
Save file under this name to local disk.
Downloads can be grouped under "tagN" so that e.g. option --Tag1 would start downloading files from that point on until next "tag1" is found. There are currently unlimited number of tag levels: tag1, tag2 and tag3, so that you can arrange your downlods hierarchially in the configuration file. For example to download all Linux files rhat you monitor, you would give option --Tag linux. To download only the NT Emacs latest binary, you would give option --Tag emacs-nt. Notice that you do not give the "level" in the option, program will find it out from the configuration file after the tag name matches.

The downloading stops at next tag of the "same level". That is, tag2 stops only at next tag2, or when upper level tag is found (tag1) or or until end of file.

    tag1: linux             # All Linux downlods under this category
        tag2: sunsite    tag2: another-name-for-this-spot
        #   List of files to download from here
        #   List of files to download from here
    tag1: emacs-binary
        tag2: emacs-nt
        tag2: xemacs-nt
        tag2: emacs
        tag2: xemacs
Extract (unpack) file after download. See also option --unpack and --no-extract The archive file, say .tar.gz will be extracted the file in current download location. (see directive lcd:)

The unpack procedure checks the contents of the archive to see if the package is correctly formed. The de facto archive format is


In the archive, all files are supposed to be stored under the proper subdirectory with version information:


"IMPORTANT:" If the archive does not have a subdirectory for all files, a subdirectory is created and all items are unpacked under it. The defualt subdirectory name in constructed from the archive name with currect date stamp in format:


If the archive name contains something that looks like a version number, the created directory will be constructed from it, instead of current date.

    package-1.43.tar.gz    =>  package-1.43
Like directive x: but extract the archive "as is", without checking content of the archive. If you know that it is ok for the archive not to include any subdirectories, use this option to suppress creation of an artificial root package-YYYY.MMDD.
This options tells to remove any previous unpack directory.

Sometimes the files in the archive are all read-only and unpacking the archive second time, after some period of time, would display

    tar: package-3.9.5/.cvsignore: Could not create file:
    Permission denied
    tar: package-3.9.5/BUGS: Could not create file:
    Permission denied

This is not a serious error, because the archive was already on disk and tar did not overwrite previous files. It might be good to inform the archive maintainer, that the files have wrong permissions. It is customary to expect that distributed packages have writable flag set for all files.


Here is list of possible error messages and how to deal with them. Turning on --debug will help to understand how program has interpreted the configuration file or command line options. Pay close attention to the generated output, because it may reveal that a regexp for a site is too lose or too tight.
ERROR {URL-HERE} Bad file descriptor
This is ``file not found error''. You have written the filename incorrectly. Double check the configuration file's line.


"Sourceforge note": To download archive files from Sourceforge requires some trickery because of the redirections and load balancers the site uses. The Sourceforge page have also undergone many changes during their existence. Due to these changes there exists an ugly hack in the program to use wget(1) to get certain infomation from the site. This could have been implemented in pure Perl, but as of now the developer hasn't had time to remove the wget(1) dependency. No doubt, this is an ironic situation to use wget(1). You you have Perl skills, go ahead and look at UrlHttGet(). UrlHttGetWget() and sen patches.

The program was initially designed to read options from one line. It is unfortunately not possible to change the program to read configuration file directives from multiple lines, e.g. by using backslashes (\) to indicate contuatinued line.


Variable "PWGET_CFG" can point to the root configuration file. The configuration file is read at startup if it exists.

    export PWGET_CFG=$HOME/conf/pwget.conf     # /bin/hash syntax
    setenv PWGET_CFG $HOME/conf/pwget.conf     # /bin/csh syntax


Not defined.


External utilities:

    wget(1)   only needed for downloads
              see BUGS AND LIMITATIONS

Non-core Perl modules from CPAN:


The following modules are loaded in run-time only if directive cnv:text is used. Otherwise these modules are not loaded:


This module is loaded in run-time only if HTTPS scheme is used:



Jari Aalto


Copyright (C) 1996-2010 Jari Aalto

This program is free software; you can redistribute and/or modify program under the terms of GNU General Public license either version 2 of the License, or (at your option) any later version.