Caching when Tumblr is down

[GoogleCodeExporter]

The caching that you have done is awesome! This help dramatically when Tumblr 
is experiencing down time. However, there are times that the cache tried to 
refresh when Tumblr is down. Hence, nothing is returned and the cache is then 
empty. It will remain empty until the cache expires. It would be nice to keep 
the old cache if Tumblr is returning zero data. What do you think?

Thanks for posting on my blog. I put quite a bit of work on top of your plugin 
and very happy with how it turned out. I plan on posting a blog entry that will 
most definitely mention your plugin. 

Original issue reported on code.google.com by joework...@gmail.com on 15 Apr 2011 at 6:37

[GoogleCodeExporter]

This can be done, but be aware that my cache system is very simple, and so can 
respond only the same exact requests made previously.

For example, if you use getPosts() to get 20 photos, and then to get 5 photos, 
the cache would not serve anything for the second request because the two 
requests are different, even if the datas from the first could answer the 
second. This kind of advanced caching would need to store informations from 
Tumblr in a SQL database, and this is far beyond the scope of this class.

Anyway, I think I can make my class to use outdated cache files when Tumblr did 
not respond.

Original comment by my.name....@gmail.com on 15 Apr 2011 at 9:56

• Changed state: Started
• Added labels: Type-Enhancement
• Removed labels: Type-Defect

[GoogleCodeExporter]

Okay, I made a first try, need testing now, so i'll wait to hear back from you 
'cause I can't force a Tumblr downtime for real testing.

You can download it here : http://static.saymonz.net/phpTumblr0.7-dev.zip

This is really bleeding edge!

How it is supposed to work?
1) On initialization, we check for the existence of a '__TumblrDown' file in 
the cache dir. If it's here and recent (created less than 10min ago), we use 
only cache, even if some files are outdated.
2) If that file don't exist, we check the avilability of the Tumblelog we want 
to read from. We assume that if the HTTP request don't end with HTTP 200 status 
code, Tumblr is down. We create the '__TumblrDown' file and use only cache.
3) When the switch 'UseOutdated' is on, caching functions will not check for 
the age of the files. We don't even try to reach Tumblr ('cause we would have 
to wait for each HTTP request to expire, not really efficient). The outdated 
files in the cache dir will not be erased as long as Tumblr stay down.

- When a getPosts() is totally or partially not in cache, getPosts returns 
false, you can however have partial results (especially when using num all or 
num > 50)
- If you request, for example, num = 20, start = 0 and then num = 20, start = 
20, you may NOT have exactly 40 posts in your results if the cache files have 
not been written at the same moment. But the 'stats' subarray hopefully don't 
lie.

Original comment by my.name....@gmail.com on 15 Apr 2011 at 11:24

[GoogleCodeExporter]

I am sorry for taking so long on getting back to you on this. I am installing 
this latest update on my blog right now. I will let it run for a bit and let 
you know how it goes. 

Original comment by joework...@gmail.com on 12 May 2011 at 11:18

[GoogleCodeExporter]

That's okay. Keep in mind that theses modifications are potentially buggy (I 
have NOT tested, just checked for syntax errors).

Original comment by my.name....@gmail.com on 12 May 2011 at 11:45

[GoogleCodeExporter]

Original comment by my.name....@gmail.com on 14 Jun 2012 at 1:02

• Changed state: Hopely-Fixed