GAE Java SDK 1.2.1 released
Of the major innovations:
1. Support for non-indexed properties of any type of database (like Text)
2. Response size increased to 10 megabytes
3. Support Thread.setContextLoader ()
Fixed various bugs for both SDK and ORM.
Be careful about (1) - as you know, without indicating to the field that it should not be indexed, the datastore builds forward and reverse indexes in 2 “bigtable” tables (without this, it is impossible to make a selection with a criterion for this field). The feature seems to reduce the amount of data in the repository, but when the field is declared as indexable in the model, the repository will not build indexes automatically (only for the data that was placed after the deploy application). The situation is the reverse, if there was already data with the indexed property, the indexes will not automatically be deleted when the property changes to non-indexable in the model. But data will no longer be added to the index.
Therefore, you will have to use remote_api and update each entry in accordance with the new model. For large data sets, this is too expensive, both in terms of update time and CPU. Why Google did not make the mechanism as described in this article (how composite indexes are built) is a mystery to me, it may have been in a great hurry.
When using this option for many applications, in addition to reducing the use of storage, it is also expected to reduce the CPU consumption required to automatically add data in 2 indexes for each property. True, there is an active discussion about this new feature.
Not yet added the innovations that are already in Python SDK 1.2.2:
1. Persistent User ID - it’s also a unique ID for Google Accounts
2. Support for key-only requests (insanely convenient for mass removal, etc.)
3. Support for deadline
in UrlFetch 4. Support in UrlFetch User- Header Agent
Well, what is really needed until it’s ready (remote_api, and most importantly local dataviewer) - it seems to be in the next version, which will be in theory soon (I hope until May 27, when they have Google I / O). Well, of course I want the rest of the features from roadmap :)
PS The eclipse plugin has also been updated, but what has changed is unknown at the time of writing.
1. Support for non-indexed properties of any type of database (like Text)
2. Response size increased to 10 megabytes
3. Support Thread.setContextLoader ()
Fixed various bugs for both SDK and ORM.
Be careful about (1) - as you know, without indicating to the field that it should not be indexed, the datastore builds forward and reverse indexes in 2 “bigtable” tables (without this, it is impossible to make a selection with a criterion for this field). The feature seems to reduce the amount of data in the repository, but when the field is declared as indexable in the model, the repository will not build indexes automatically (only for the data that was placed after the deploy application). The situation is the reverse, if there was already data with the indexed property, the indexes will not automatically be deleted when the property changes to non-indexable in the model. But data will no longer be added to the index.
Therefore, you will have to use remote_api and update each entry in accordance with the new model. For large data sets, this is too expensive, both in terms of update time and CPU. Why Google did not make the mechanism as described in this article (how composite indexes are built) is a mystery to me, it may have been in a great hurry.
When using this option for many applications, in addition to reducing the use of storage, it is also expected to reduce the CPU consumption required to automatically add data in 2 indexes for each property. True, there is an active discussion about this new feature.
Not yet added the innovations that are already in Python SDK 1.2.2:
1. Persistent User ID - it’s also a unique ID for Google Accounts
2. Support for key-only requests (insanely convenient for mass removal, etc.)
3. Support for deadline
in UrlFetch 4. Support in UrlFetch User- Header Agent
Well, what is really needed until it’s ready (remote_api, and most importantly local dataviewer) - it seems to be in the next version, which will be in theory soon (I hope until May 27, when they have Google I / O). Well, of course I want the rest of the features from roadmap :)
PS The eclipse plugin has also been updated, but what has changed is unknown at the time of writing.