- rawdog 2.11

Avoid a crash when a feed's URL is changed and expiry is done on the
same run.

Encode dates correctly in non-ASCII locales (reported by Damjan
Georgievski).

Strengthen the warning in PLUGINS about the effects of overriding
output_write_files (suggested by Virgil Bucoci).

Add the state directory to sys.path, so you can put modules that plugins
need in your ~/.rawdog (suggested by Stuart Langridge).

When adding a feed, check that it isn't already present in the config
file (suggested by Stuart Langridge).

Add --no-lock-wait option to make rawdog exit silently if it can't lock
the state file (i.e. if there's already a rawdog running).

Update to the latest feedparser, which fixes an encoding bug with Python
2.5, among various other stuff (reported by Paul Tomblin, Tim Bishop and
Joseph Reagle).

Handle the author_detail fields being None.

- rawdog 2.10

Work around a feedparser bug (returning a detail without a type field
for posts with embedded SVG).

Pull in most of the changes from feedparser 4.1.

Fix a bug that stopped rawdog from working properly when no locale
information was present in the environment, or on versions of Python
without locale.getpreferredencoding() (reported by Michael Watkins).

Add --remove option to remove a feed from the config file (suggested by
Wolfram Sieber).

Produce a more useful error message when $HOME isn't set (reported by
Wolfram Sieber).

Fix a bug in the expiry code: if you were using keepmin, it could expire
articles that were no longer current but should be kept.

Clean up the example config file a bit.

- rawdog 2.9

Fix a documentation bug about time formats (reported by Tim Bishop).

Fix a text-handling problem related to the locale changes (patch from
Samuel Hym).

Fix use of the "A-IM: feed" header in HTTP requests. A previous upstream
change to feedparser had modified it so that it always sent this header,
which results in a subtle rawdog bug: if a feed returns a partial result
(226) and then has no changes for a long time, rawdog can expire
articles which should still be "current" in the feed. This version adds
a "keepmin" option which make a minimum number of articles be kept for
each feed; this should avoid expiring articles that are still current.
If you want the old behaviour, you can set "keepmin" to 0, in which case
rawdog won't send the "A-IM: feed" header in its requests. rawdog also
won't send that header if "currentonly" is set to true, since in that
case the current set of articles is all rawdog cares about. (See
<http://www.intertwingly.net/blog/2006/04/29/Now-you-see-it> for Sam
Ruby's discussion of the same problem in Planet.)

If the author's name is given as the empty string, fall back to the
email address, URL or "author".

Change the labels in the feed information table to "Last fetched" and
"Next fetched after", to match what rawdog actually does with the times
it stores (reported by D. Stussy).

- rawdog 2.8

Fix authentication support -- feedparser now supports Basic and Digest
authentication internally, but it needed tweaking to make it useful for
rawdog (reported by Tim Bishop).

- rawdog 2.7

Make feedfinder smarter about trying to find the preferred type of feed
(patch from Decklin Foster).

Add a plugin hook to let you modify mx.Tidy options (suggested by Jon
Lasser).

Work correctly if the threading module isn't available (patch from Jack
Diederich).

Update to feedparser 4.0.2, which includes some of our patches and fixes
an unclear license notice (reported by Jason Diamond, Joe Wreschnig and
Decklin Foster).

Fix a feedparser bug that caused things preceding shorttags to be
duplicated when sanitising HTML.

Set the locale correctly when rawdog starts up (patch from Samuel Hym).

- rawdog 2.6

Allow maxage to be set per feed (patch from Craig Allen).

Support feeddefaults with no options on the same line, as used in the
sample config file (reported by asher).

- rawdog 2.5

Ensure that all the strings in entry_info are in Unicode form, to make
it easier for plugins to deal with them.

Fix a feedparser bug that was breaking feeds which includes itunes
elements (reported by James Cameron).

Make feedparser handle content types and modes in atom:content correctly
(reported by David Dorward).

Make feedparser handle the new elements in Atom 1.0 (patch from Decklin
Foster).

Remove some unnecessary imports found by pyflakes.

Add output_sorted_filter and output_write_files hooks, deprecating
the output_write hook (which wasn't very useful originally, and isn't
used by any of the plugins I've been sent). Restructure the "write" code
so that it should be far easier to write custom output plugins: there
are several new methods on Rawdog for doing different bits of the write
process.

When selecting articles to display, don't assume they're sorted in date
order (a plugin might have done something different).

Don't write an extra newline at the end of the output file (i.e. use
f.write rather than print >>f), and be more careful about encoding when
writing output to stdout.

Provide arbitrary persistent storage for plugins via a
get_plugin_storage method on Rawdog (suggested by BAM).

Add -N option to avoid locking the state file, which may be useful if
you're on an OS or filesystem that doesn't support locks (suggested by
Andy Dustman).

If RAWDOG_PROFILE is set as an environment variable, rawdog will run
under the Python profiler.

Make some minor performance improvements.

Change the "Error parsing feed" message to "Error fetching or parsing
feed", since it really just indicates an error somewhere within
feedparser (reported by Fred Barnes).

Add support for using multiple threads when fetching feeds, which makes
updates go much faster if you've got lots of feeds. (The state-updating
part of the update is still done sequentially, since parallelising it
would mean adding lots of locking and making the code very messy.) To
use this, set "numthreads" to be greater than 0 in your config file.
Since it changes the semantics of one of the plugin hooks, it's off by
default.

Update the GPL and LGPL headers to include the FSF's new address
(reported by Decklin Foster).

- rawdog 2.4

Provide guid in item templates (suggested by Rick van Rein).

Update article-added dates correctly when "currentonly true" is used
(reported by Rick van Rein).

Clarify description of -c in README and man page (reported by Rick van
Rein).

If you return false from an output_items_heading function, then disable
DayWriter (suggested by Ian Glover).

Fix description of article_seen in PLUGINS (reported by Steve Atwell).

Escape odd characters in links and guids, and add a sanity check that'll
trip if non-ASCII somehow makes it to the output (reported by
TheCrypto).

- rawdog 2.3

Make the id= parameter work correctly (patch from Jon Nelson).

- rawdog 2.2

Add "feeddefaults" statement to specify default feed options.

Update feeds list from the config file whenever rawdog runs, rather than
just when doing an update (reported by Decklin Foster).

Reload the config files after -a, so that "rawdog -a URL -u" has the
expected behaviour (reported by Decklin Foster).

Add "define" statement and "define_X" feed option to allow the user to
define extra strings for the template; you can use this, for example, to
select classes for groups of feeds, generate different HTML for
different sorts of feeds, or set the title in different pages generated
from the same template (suggested by Decklin Foster).

Fix a logic error in the _raw changes to feedparser: if a feed didn't
specify its encoding but contained non-ASCII characters, rawdog will
now try to parse it as UTF-8 (which it should be) and, failing that,
as ISO-8859-1 (in case it just contains non-UTF-8 junk).

Don't print the "state file may be corrupt" error if the user hits
Ctrl-C while rawdog's loading it.

Add support for extending rawdog with plugin modules; see the "PLUGINS"
file for more information.

Make "verbose true" work in the config file.

Provide __author__ in items, for use in feeds that support that (patch
from Decklin Foster).

Fix conditional template expansion (patch from Decklin Foster).

Add "blocklevelhtml" statement to disable the "<p>" workaround for
non-block-level HTML; this may be useful if you have a plugin that is
doing different HTML sanitisation, or if your template already forces a
block-level element around article descriptions.

Fix -l for feeds with non-ASCII characters in their titles.

Provide human-readable __feed_id__ in items (patch from David
Durschlag), and add feed-whatevername class to the default item
template; this should make it somewhat easier to add per-feed styles.

Handle feeds that are local files correctly, and handle file: URLs in
feedparser (reported by Chris Niekel).

Allow feed arguments to be given on indented lines after the "feed" or
"feeddefaults" lines; this makes it possible to have spaces in feed
arguments.

Add a meta element to the default template to stop search engines
indexing rawdog pages (patch from Rick van Rein).

Add new feeds at the end of the config file rather than before the first
feed line (patch from Decklin Foster).

- rawdog 2.1

Fix a character encoding problem with format=text feeds.

Add proxyuser and proxypassword options for feeds, so that you can use
per-feed proxies requiring HTTP Basic authentication (patch from Jon
Nelson).

Add a manual page (written by Decklin Foster).

Remove extraneous #! line from feedparser.py (reported by Decklin
Foster).

Update an article's modified date when a new version of it is seen
(reported by Decklin Foster).

Support nested ifs in templates (patch from David Durschlag), and add
__else__.

Make the README file list all the options that rawdog now supports
(reported by David Durschlag).

Make --verbose work even if it's specified after an action (reported by
Dan Noe and David Durschlag).

- rawdog 2.0

Update to feedparser 3.3. This meant reworking some of rawdog's
internals; state files from old versions will no longer work with rawdog
2.0 (and external programs that manipulate rawdog state files will also
be broken). The new feedparser provides a much nicer API, and is
significantly more robust; several feeds that previously caused
feedparser internal errors or Python segfaults now work fine.

Add an --upgrade option to import state from rawdog 1.x state files into
rawdog 2.x. To upgrade from 1.x to 2.x, you'll need to perform the
following steps after installing the new rawdog:
- cp -R ~/.rawdog ~/.rawdog-old
- rm ~/.rawdog/state
- rawdog -u
- rawdog --upgrade ~/.rawdog-old ~/.rawdog (to copy the state)
- rawdog -w
- rm -r ~/.rawdog-old (once you're happy with the new version)

Keep track of a version number in the state file, and complain if you
use a state file from an incompatible version.

Remove support for the old option syntax ("rawdog update write").
Remove workarounds for early 1.x state file versions.

Save the state file in the binary pickle format, and use cPickle instead
of pickle so it can be read and written more rapidly.

Add hideduplicates and allowduplicates options to attempt to hide
duplicate articles (based on patch from Grant Edwards).

Fix a bug when sorting feeds with no titles (found by Joseph Reagle).

Write the updated state file more safely, to reduce the chance that
it'll be damaged or truncated if something goes wrong while it's being
written (requested by Tim Bishop).

Include feedfinder, and add a -a|--add option to add a feed to the
config file.

Correctly handle dates with timezones specified in non-UTC locales
(reported by Paul Tomblin and Jon Lasser).

When a feed's URL changes, as indicated by a permanent HTTP redirect,
automatically update the config file and state.

- rawdog 1.13

Handle OverflowError with parsed dates (patch from Matthew Scott).

- rawdog 1.12

Add "sortbyfeeddate" option for planet pages (requested by David
Dorward).

Add "currentonly" option (patch from Chris Cutler).

Handle nested CDATA blocks in feed XML and HTML correctly in feedparser.

- rawdog 1.11

Add __num_items__ and __num_feeds__ to the page template, and __url__ to
the item template (patch from Chris Cutler).

Add "daysections" and "timesections" options to control whether to split
items up by day and time (based on patch from Chris Cutler).

Add "tidyhtml" option to use mx.Tidy to clean feed-provided HTML.

Remove the <p> wrapping __description__ from the default item template,
and make rawdog add <p>...</p> around the description only if it doesn't
start with a block-level element (which isn't perfect, but covers the
majority of problem cases). If you have a custom item template and want
rawdog to generate a better approximation to valid HTML, you should
change "<p>__description__</p>" to "__description__".

HTML metacharacters in links are now encoded correctly in generated
HTML ("foo?a=b&c=d" as "foo?a=b&amp;c=d").

Content type selection is now performed for all elements returned from
the feed, since some Blogger v5 feeds cause feedparser to return
multiple versions of the title and link (reported by Eric Cronin).

- rawdog 1.10

Add "ignoretimeouts" option to silently ignore timeout errors.

Fix SSL and socket timeouts on Python 2.3 (reported by Tim Bishop).

Fix entity encoding problem with HTML sanitisation that was causing
rawdog to throw an exception upon writing with feeds containing
non-US-ASCII characters in attribute values (reported by David Dorward,
Dmitry Mark and Steve Pomeroy).

Include MANIFEST.in in the distribution (reported by Chris Cutler).

- rawdog 1.9

Add "clear: both;" to item, time and date styles, so that items with
floated images in don't extend into the items below them.

Changed how rawdog selects the feeds to update; --verbose now shows
only the feeds being updated.

rawdog now uses feedparser 2.7.6, which adds date parsing and limited
sanitisation of feed-provided HTML; I've removed rawdog's own
date-parsing (including iso8601.py) and relative-link-fixing code in
favour of the more-capable feedparser equivalents.

The persister module in rawdoglib is now licensed under the LGPL
(requested by Giles Radford).

Made the error messages that listed the state dir reflect the -b
setting (patch from Antonin Kral).

Treat empty titles, links or descriptions as if they weren't supplied at
all, to cope with broken feeds that specify "<title></title>" (patch
from Michael Leuchtenburg).

Make the expiry age configurable; previously it was hard-wired to 24
hours. Setting this to a larger value is useful if you want to have a
page covering more than a day's feeds.

Time specifications in the config file can now include a unit; if no
unit is specified it'll default to minutes or seconds as appropriate to
maintain compatibility with old config files. Boolean values can now be
specified as "true" or "false" (or "1" or "0" for backwards
compatibility). rawdog now gives useful errors rather than Python
exceptions for bad values. (Based on suggestions by Tero Karvinen.)

Added datetimeformat option so that you can display feed and article
times differently from the day and time headings, and added some
examples including ISO 8601 format to the config file (patch from Tero
Karvinen).

Forcing a feed to be updated with -f now clears its ETag and
Last-Modified, so it should always be refetched from the server.

Short-form XML tags in RSS (<description/>) are now handled correctly.

Numeric entities in RSS encoded content are now handled correctly.

- rawdog 1.8

Add format=text feed option to handle broken feeds that make their
descriptions unescaped text.

Add __hash__ and unlinked titles to item templates, so that you can use
multiple config files to build a summary list of item titles (for use in
the Mozilla sidebar, for instance). (Requested by David Dorward.)

Add the --verbose argument (and the "verbose" option to match); this
makes rawdog show what it's doing while it's running.

Add an "include" statement in config files that can be used to include
another config file.

Add feed options to select proxies (contributed by Neil Padgen). This is
straightforward for Python 2.3, but 2.2's urllib2 has a bug which
prevents ProxyHandlers from working; I've added a workaround for now.

- rawdog 1.7

Fix code in iso8601.py that caused a warning with Python 2.3.

- rawdog 1.6

Config file lines are now split on arbitary strings of whitespace, not
just single spaces (reported by Joseph Reagle).

Include a link to the rawdog home page in the default template.

Fix the --dir argument: -d worked fine, but the getopt call was missing
an "=" (reported by Gregory Margo).

Relative links (href and src attributes) in feed-provided HTML are now
made absolute in the output. (The feed validator will complain about
feeds with relative links in, but there are quite a few out there.)

Item templates are now supported, making it easier to customise item
appearance (requested by a number of users, including Giles Radford and
David Dorward). In particular, note that __feed_hash__ can be used
to apply a CSS style to a particular feed.

Simple conditions are supported in templates: __if_x__ .. __endif__ only
expands to its contents if x is not empty. These conditions cannot be
nested.

PyXML's iso8601 module is now included so that rawdog can parse dates in
feeds.

- rawdog 1.5

Remove some debugging code that broke timeouts.

- rawdog 1.4

Fix option-compatibility code (reported by BAM).

Add HTTP basic authentication support (which means modifying feedparser
again).

Print a more useful error if the statefile can't be read.

- rawdog 1.3

Reverted the "retry immediately" behaviour from 1.2, since it causes
denied or broken feeds to get checked every time rawdog is run.

Updated feedparser to 2.5.3, which now returns the XML encoding used.
rawdog uses this information to convert all incoming items into Unicode,
so multiple encodings are now handled correctly. Non-ASCII characters
are encoded using HTML numeric character references (since this allows
me to leave the HTML charset as ISO-8859-1; it's non-trivial to get
Apache to serve arbitrary HTML files with the right Content-Type,
and using <meta http-equiv="Content-Type"...> won't override HTTP
headers).

Use standard option syntax (i.e. "--update --write" instead of "update
write").  The old syntax will be supported until 2.0.

Error output from reading the config file and from --update now goes to
stderr instead of stdout.

Made the socket timeout configurable (which also means the included copy
of feedparser isn't modified any more).

Added --config option to read an additional config file; this lets you
have multiple output files with different options.

Allow "outputfile -" to write the output to stdout; useful if you want
to have cron mail the output to you rather than putting it on a web
page.

Added --show-template option to show the template currently in use (so
you can customise it yourself), and "template" config option to allow
the user to specify their own template.

Added --dir option for people who want two lots of rawdog state (for two
sets of feeds, for instance).

Added "maxage" config option for people who want "only items added in
the last hour", and made it possible to disable maxarticles by setting
it to 0.

- rawdog 1.2

Updated feedparser to 2.5.2, which fixes a bug that was making rawdog
handle content incorrectly in Echo feeds, handles more content encoding
methods, and returns HTTP status codes. (I've applied a small patch to
correct handling of some Echo feeds.)

Added useful messages for different HTTP status codes and HTTP timeouts.
Since rawdog reads a config file, it can't automatically update
redirected feeds, but it will now tell you about them.  Note that for
"fatal" errors (anything except a 2xx response or a redirect), rawdog
will now retry the feed next time it's run.

Prefer "content" over "content_encoded", and fall back correctly if no
useful "content" is found.

- rawdog 1.1

rawdog now preserves the ordering of articles in the RSS when a group of
articles are added at the same time.

Updated rawdog URL in setup.py, since it now has a web page.

Updated rssparser to feedparser 2.4, and added very preliminary support
for the "content" element it can return (for Echo feeds).

- rawdog 1.0

Initial stable release.