Before, I would get random errors about shards not being available. Erik Rose told me that this is because I made an index, but then didn't wait for it to propogate to the shards. He gave me a line of code to wait for the sync. I noticed I had to add it in several places, always right after a call to `setup_indexes()`. So I made that function do that. I also removed the refresh calls right after any `setup_index()` calls, because refresh is called in that function. The old way of working around this was to sleep a bit during refreshes. That isn't as good as this way, so I nuked the old way, including the docs for it.
This updates the codebase to use the pyelasticsearch-based elasticutils from master tip. Generally the changes are pretty minor. Most of them are to handle the pyelasticsearch exceptions rather than the pyes ones. A few are because some of the API bits changed in elasticutils. Then there's some very minor cleanup.
This updates to the master tip of elasticutils which uses pyelasticasearch. It ditches pyes. It adds pyelasticsearch and its requirements. It fails tests because the code changes haven't been made, yet.
Fill the graph with data at every day along the X axis, instead of allowing missing data to be interpolated. Care is taken to make sure that these extra zero points align to the spacing of the rest of the points (24 hours per point). Other, related changes: - If start or end is not specified, they have default values now. - The infinity date selector was replaced with a 90 day button. - The one day selector now hides the graph. The zero fill algorithm benefits from having non-None start and end times, and these changes help with that.
This updates playdoh-lib: ffb6542 Merge pull request #29 from peterbe/revert-django-browserid-0.7.1 0b784af Revert "upgrade to django-browserid==0.7.1" 5f367d6 Merge pull request #27 from peterbe/upgrade-django-browserid d1d4eed upgrade to django 1.4.3, r=mkelly a78e698 upgrade to django-browserid==0.7.1 8508fd6 Merge pull request #25 from willkg/ordereddict-fix 9ea2e90 Add ordereddict d357594 upgrading funfactory for latest urlencode() fix, r=willkg 22bfbe9 Merge pull request #23 from willkg/celery-2.5-update 275892b Add celery 2.5.5 and friends 92de02f Remove celery 2.1.4, djcelery 2.1.4 and amqplib 0.8
This makes the graph lines and the graph data points line up. This only changes flot's expectations to match what elastic is ending it. When elasticsearch returns histogram data with a period of 1 day (like we are doing on the histogram) the timestamps it returns have the hours of a UTC midnight date, regardless of what timezone the aggregate set represents. This means that when the server is working in Pacific time, it aggregates a Pacific day worth of data, sums it up, and then returns it labeled with a timestamp. This timestamp, when output, is the midnight of a UTC day (calculating the hours past midnight only returns 0 if you assume UTC), even if localtime on the server is Pacific. Even if we change to forcing a particular timezone in the future, this will still be true, and elasticwill still label this data with UTC, regardless of the input. Because of this, flot needs to be told the data is UTC, not browser local time.
Vendor is a gitsubmodule, which means we should be obeying git's idea of what commit it should be on. We should not checkout out a particular branch, because this can cause changes to come in to early, such as the celery 2.6 upgrade, which caused bug 805020. Additionally, I added --recursive to the submodule checkout in the main repo, since we aren't doing this explicitly in vendor anymore.
Fill the graph with data at every day along the X axis, instead of allowing missing data to be interpolated. Care is taken to make sure that these extra zero points align to the spacing of the rest of the points (24 hours per point). Other minor, related changes: - If start or end is not specified, they have default values now. - The infinity date selector was replaced with a 90 day button. - The one day selector now hides the graph. - Due to the fixed, relative time frame, tests have been changed to use relative dates. The zero fill algorithm benefits from having non-None start and end times, and these changes help with that.
This updates elasticutils to master tip. The changes are mostly cosmetic: * 1578b37 - Issue 85. Fix IndexMissingException. * 7dcc034 - Issue 84. Fix get_indexes/get_doctypes silliness * cd76217 - Fix v0.5/v0.6 issues in changelog * 675419f - Update hacking_howto about django reqs. * 374d224 - Fix boosted text_phrase queries * a478db3 - Update release date for 0.5; add TOC * 292e18d - Add Giorgos to CONTRIBUTORS list * afbd458 - Update CHANGELOG
This adds a checkbox and a input box to the survey form. The checkbox is to indicate the user wants to give an email address, and the input field is for the email address, and only shows up after the box has been checked. This email address is stored in a seperate table with a foreign key, for possible security benefits down the road.
I spent some time extracting the sampledata application out as a separate project so we could use it in Kitsune (and other places as well). This switches fjord over to using that. One nice thing eadred has is the ability to pass parameters from the command line to the generate_sampledata functions. This allows us to easily allow for different kinds of sample data generation: reproducible data set for contributors, randomized large data set for testing the dashboard, icky data set for testing security and unicode handling, ...
* add app settings view so we can see server configuration * add celery settings view so we can see celery configuration * add wsgi env view so we can see the wsgi environment (though we haven't needed this, yet) Note: The template is a Django template, so it uses Django template stuff. Also, I lifted this from Kitsune so it's pretty much the same code that's in Kitsune with some very minor cleanup.