Removed limits on zigzag merge-join requests. Now, for most complex queries, composite indexes are not needed, which greatly increased the amount of storage space. More details can be viewed from the presentation on Google IO - Next gen queries
Maximum trash size for TaskQueue increased to 100
Added parameter "image quality" for operations with the Image API
For Java: Adding java.net.InetAddress and some interfaces and abstract classes from javax.xml.soap to the white list.
For Python: Support for builtin-handlers is a great feature that allows you to enable some SDK functionality (for example, remote_api, mapreduce, etc) right out of the box, with a single hand movement. Below is an example to enable appstats and mapreduce. Read more about builtin and includes
builtins: - mapreduce: on - appstats: on
Changes in admin production:
Instances console - now you can see how many of them are running for the application, as well as see their parameters (QPS, Latency, Memory)
You can run tasks directly from the admin
With the built-in datastore_admin enabled, you can delete all (or part of) the data from a specific table automatically using mapreduce for this purpose. However, it must be remembered that it is resource intensive, and you can get to the limits
')
SDK analysis - a certain Matcher API appeared for python - however its purpose is extremely incomprehensible (maybe part of the future full-text-search (?), Judging from stub analysis), and it looks like the API itself is still in deep development. Also, it seems, we will soon expect asynchronous calls to the repository, similar to URLFetch.