twitpic seems to prevent the retrieval of the absolute image url via the api.

a possible solution: parse the html of the twitpic page showing the image.

quite self explanatory - enabling the user to order posts by authir.

go grab this ticket & tag new posts with a tag, e.g. @sven_kr -> #sven_kr

Every few weeks I see this error in the logs:

/home/5v3n/.rvm/gems/ruby-1.9.2-p180@tweetlr/gems/tweetlr-0.1.3/lib/tweetlr.rb:251:inrescue in http_get': incompatible character encodings: ASCII-8BIT and UTF-8 (Encoding::CompatibilityError)`

Which could be either a response in utf-8 that could not be read, or some "hidden utf-8 code block" in the ruby file that wasn't executed before.

veeeery rare, but hurst nonetheless: twitter hostname resolving error:

.../ruby-1.9.2-p290@tweetlr/bundler/gems/tweetlr-05ed2f5739c6/lib/tweetlr.rb:253:in 'perform': Couldn't resolve host name (Curl::Err::HostResolutionError) from /home/5v3n/.rvm/gems/ruby-1.9.2-p290@tweetlr/bundler/gems/tweetlr-05ed2f5739c6/lib/tweetlr.rb:253:in 'http_get' from /home/5v3n/.rvm/gems/ruby-1.9.2-p290@tweetlr/bundler/gems/tweetlr-05ed2f5739c6/lib/tweetlr.rb:116:in 'lazy_search_twitter'

The daemons gem redirects all output to a log file. Plus preventing Logger to write to an own file itself.

Right now, all log calls are replaced w/ puts to get around this "feature". Here's a way to use Logger for that - just redirect it to stdout & we should be fine.

http://stackoverflow.com/questions/224512/redirect-the-puts-command-output-to-a-log-file

Requested by @LasseL - who thinks it's a good idea to be able to collect all the pictures in a single download so you can do a giant poster.

I agreed, although I'd prefer a Moocards stickerbook :-).

just found this post from the tumblr staff (http://staff.tumblr.com/post/6074690226/soundcloud). tumblr is now a sharing feature at soundcloud and posting soundcloud tracks as audio posts on tumblr is now possible.

with the soundcloud apps, you can record, edit metadata, publish to soundcloud and share to twitter in one go.

may be far fetched, but with pictures, video and audio, you had something like a decentralized reporting network in place...

the quick & dirty way I hacked the tests w/ live calls is kind of embarrassing ;-)

When a tweetlr hits the tumblr photo quota wall, we currently save the tweet id in the TweetlrConfiguration.since_id field. If we would only save it for successful, non-quota-impeded tumblr photo posts we have a chance of catching/posting the pics the next day.

right now, tweetlr grew into a direction where the "a tiny little script needs no decoupling"-state is not acceptable any more.

plan: decouple functionality & use a highly cohesive, module based approach

The API v1 will be retired - we have to move to v1.1!

see: https://dev.twitter.com/blog/api-v1-retirement-final-dates

Docs:

https://dev.twitter.com/docs/api/1.1

videos (at least youtube, vimeo would also be nice) => post as video
links (i know, probably not as easy as it sounds, since pics and videos are basically links in the first place too...) => post as link

Inconsistent behavior:

while the general tumblr tag works:

http://www.tumblr.com/tagged/sven_kr

the blog specific one does not:

http://hktbl.tumblr.com/tagged/sven_kr

can't really see the cause...

should be implemented since it's standard in some clients.

see http://groups.google.com/group/plixi/web/fetch-photos-from-url?pli=1 for the piece of cake api call that's needed.

basic auth is sufficiently secure for servers / hosted tweetlrs. but if you're using a laptop to stream your event in a free wifi, it's probably more cozy to use cookie auth.

infinite recursive loop ahead!

same is true if the tweetlr-tumblr is mentionen in a tweet with the corresponding hashtag (no endless loop, but annoying)

tweetlr gives following error, with or without arguments:

/Library/Ruby/Gems/1.8/gems/tweetlr-0.0.6/bin/tweetlr:7: undefined method require_relative' for main:Object (NoMethodError) from /usr/bin/tweetlr:19:inload'
from /usr/bin/tweetlr:19

any ideas?

embed.ly doesn't support eyeem - we need a handcrafted crawler. see lib/processors/photo_services.rb and spec/processors/photo_services_spec.rb for examples of other services e.g. path and foursquare.

tweets are lost within hours. but they are just 140 bytes + metadata. since they are grabbed and analyzed by tweetlr anyways, why not dump them in a simple text file? or a ultra-light database table for later export?

add posterous support

Currently RT & quoted retweets are handled - we need MT in addition to it.

currently, the first link is checked if it's a link to a photo or not. So we miss all pics that have another link in front.

Currently, all searches are "OR" connected. Using the twitter search syntax, you could do "AND" and "OR" searches for terms...

mixed retries and retries_left when handling http calls - so failure retry fails. d'oh!'

Right now the primary tumblelog is used automatically. To put content to a secondary blog we need to

1. introduce the configuration option
2. use this option if set to select the desired blog
3. update the "tweetlr as a service" webinterface accordingly (has to be done by @5v3n)

see http://www.tumblr.com/docs/en/api

currently, only the url aprameter of embed.ly is used.

with 500px.com / http://www.eyeem.com , you need to use embed.ly's preview/objectify function & extract the first image in the response:

regular: http://embed.ly/docs/explore/oembed?url=http%3A%2F%2Ft.co%2FsEtPUbn

objectify: http://api.embed.ly/2/objectify?url=http://t.co/sEtPUbn&maxwidth=500

motivation:

• OAuth-Support replaces HTTP basic auth
• that's enough already ;-)

see http://www.tumblr.com/docs/en/api/v2 for details.

EDIT: Promoted to bug since... TUMBLR SHUT DOWN V1 API!!!111!!!

http://developers.tumblr.com/post/28557510444/welcome-to-the-official-tumblr-developers-blog

Means: no tweetlr will work until this one is fixed!

Alternatively: we could move to posterous. No quota there...

use embed.ly for all urls that could not be handled by tweetlr directly.

Solution: ignore tweets with "Photo: "

the picplz parser crashes due to nil objects - check for them to increase stability / continuity

Seems like Logger does not work when running tweetlr w/ the Daemons gem. This is due to Daemons redirecting all output to STDOUT / STDERR to a logfile itself.

Solution could be: do not use Logger but puts to STDOUT / STDERR. Which feels kind of sucky...

For example #hashtag1+#hashtag2 -> %23hashtag1%2B%23hashtag2 and searchterm1 + searchterm2 -> searchterm1%20+%20searchterm2

noticed that tumblr is not using https by default (neither for logging in or creating accounts) and was wondering how tweetlr is handling this. is my email-adress / password duo sent through the web in plaintext every five minutes?

right now, you have to manually adjust the twitter timestamp / since_id each time you restart the crawler. persisting it (e.g. in a file) would take much ops hassle away.

same is true for any tweet with a link to a html page that has an image in it - guess embed.ly is configured too loosely, better remove "thumbnail fallback".

Update: digiwomen-tumblr affected too, high priority task

we persist tweets.last['id'] which appears to be the oldest tweet collected. we have to use tweets.first['id']. a little counter-intuitive... but that's the tweet id we need.