Griatch
32b069c5fc
Added the ability to store Sets in Attributes (using a PackedSet object type).
2012-10-20 12:10:03 +02:00
Griatch
4830560ac5
Added better error checking for when trying to set an attribute on an object stored in an attribute but which after retrieved and cached was deleted. The fix doesn't fully fix the problem of stale caching but does give more information. Related to Issue 295.
...
Also fixed a curiously phrased import.
2012-10-14 19:29:56 +02:00
Griatch
a347252407
Added the ability to clean the attribute_cache on a per-object bases using obj.flush_attr_cache().
2012-09-29 16:02:43 +02:00
Griatch
9a3562148a
Reverting indexing of objattribute.db_value; it is a highly db-incompatible change, and also lacking the possibility to create unique indices with unclear results.
2012-09-28 09:08:43 +02:00
Griatch
b0ac916b9a
Obs-run migrations! Added an index to the ObjAttribute's values.
2012-09-28 00:14:29 +02:00
Griatch
0dae03156c
Some optimizations, cleanup and a few bugfixes. Just changing a spurious property retrieval in the typeclass removed an extra, pointless database query.
2012-09-18 01:03:35 +02:00
Griatch
ffcf4b3c2f
Added cache-resyncing to ProcPool. This makes sure to update all affected object caches
...
whenever the subprocess returns (this is potentially not good enough for long-running scripts,
will have to ponder that one). Made ProcPool work with MySQL (where it works much better). Tested and fixed many
small bugs.
2012-09-03 01:11:14 +02:00
Griatch
f68523cc22
Minor optimizations and some fixes to the dummyrunner.
2012-08-22 22:34:43 +02:00
Griatch
cc88d38ab6
Some optimizations towards speeding up getting the contents of a location.
2012-08-19 11:45:13 +02:00
Griatch
1dee271fc7
Removed a bug in property cache that disabled the cache if the property was null (e.g for player lookups)
2012-08-19 11:15:22 +02:00
Griatch
7dcfdc4d4b
Fixed a minor typo in docstring for ObjectDB.__db_get: To view all Attributes one uses obj.db.all, not obj.db.all().
2012-06-20 23:22:26 +02:00
Griatch
3d24ee2242
Added a secure_attr() method to the base typeclass. This allows for system to access attributes with full access control (since it requires an accessing_obj as argument). This relies on (by default unused) lock types "readattr" (reading/listing the attr value), "editattr" (editing/deleting the attribute) and finally "createattr" for creating a new attribute. The first two checks are done directly on the Attribute class whereas the createattr check is done on the TypedObject on which the Attribute is to be stored.
2012-06-10 21:46:00 +02:00
Griatch
2dba8ad547
Changed attribute-caching to use a string as key rather than the attribute-object itself. The latter caused the database to be invoked in order to create the hash over and over.
2012-04-28 14:47:11 +02:00
Griatch
3091587e33
Added a timeout to the attribute caching; the system will now clean cache at regular intervals once it pass a certain size defined in settings.
2012-04-28 00:37:36 +02:00
Griatch
e3ce0a7933
Some fixes to the caching, particularly in set_attribute.
2012-04-27 00:03:31 +02:00
Griatch
9d970ea7c5
Added full caching to all attributes on a per-object level. This speeds up attribute loading considerably.
2012-04-26 21:39:16 +02:00
Griatch
6e08c011a1
Further caching and optimization, making some operations noticeable faster in the end.
2012-04-26 17:47:25 +02:00
Griatch
1a6ef5d983
Added more caching to channelhandler as well as players in order to cut back on unnecessary database calls.
2012-04-26 13:38:34 +02:00
Griatch
8c3b49e704
Changed how the Typeclass system returns errors. Instead of echoing typeclass erros to the MUD-info channel (which is not only not only very spammy for everyone but also very hard to make clean so as to avoid recursion at a stage of typeclass failing), the system instead stores a property on itself called 'typeclass_last_errmsg' that holds eventual errors. This means that the task of reporting errors does not fall on the typeclass system itself but on the calling methods, as it should be. So src.utils.create.create_* functions now takes a new optional keyword "report_to" that holds an object to receive errors. If this keyword is given, the function msg():es that object with the error and returns None as before. If report_to is not set however, the create_* methods now return an Exception containing the error text. All default commands have been changed to accomodate for this behaviour, which allows for much more control over errors.
...
Also, the default ADMIN_MEDIA static files changed location in Django 1.4. The initial_setup function now accounts for this.
2012-04-21 16:15:37 +02:00
Griatch
464aa8ca9e
Added the ability to clear an object from the global cache. This is rarely
...
needed (and can be potentially dangerous if the object depends on certain
startup methods to run and/or holds temporary attributes on themselves -
these will all be lost due to a new instance being created. It is hoever
necessary when it comes to renaming Exits - since the Command on the exit
must then change name too, recaching the Exit will also update the command.
Resolves issue 223.
2012-04-15 21:46:43 +02:00
Griatch
c8df141e89
Further cleanup of source; making class methods _private for clarity in the API.
2012-03-31 15:09:22 +02:00
Griatch
fc156b5a54
Continuing work on API, making local-use class properties _private to make things easier to explore.
2012-03-31 13:06:29 +02:00
Griatch
c0322c9eae
Cleaning some unnecessary whitespace, overall cleanup of various source codes.
2012-03-30 23:47:22 +02:00
Griatch
d4c97d7df8
Fixed a caching issue with TypedAttributes.
2012-03-29 20:30:35 +02:00
Griatch
82a10903d1
Added more comments to help exploration through the ev interface.
2012-03-29 19:42:08 +02:00
Griatch
4398d42360
Added new templates to gamesrc/*/examples. The old base* modules still in place. Some cleanup of the API.
2012-03-25 16:35:22 +02:00
Griatch
6a78fdafcb
Removed a debug-print statement.
2012-03-13 01:39:30 +01:00
Griatch
ca280af659
Fixed cache issues with @type and @name commands that didn't properly update the cache when using them.
2012-03-13 01:37:28 +01:00
Griatch
9660dd6656
Added caching to some more object properties.
2012-02-26 12:43:16 +01:00
Griatch
24c4791e9e
Fixed a bug with the updating of the caching.
2012-02-26 01:10:20 +01:00
Griatch
1fa5791e19
Added aggressive caching for common typeclassed variables.
2012-02-25 23:37:50 +01:00
Griatch
9733a43a16
Fixed the issues with assigning to nested attribute-lists/dicts "in-situ", e.g. obj.db.mylist[1][2] = val. This now works as it should. I'm still not sure this behaviour is worth the fact that returning mylist from the attribute will actually return a nested structure of PackedLists/Dicts, and these will continue to save to the database whenever they are updated, also if not used in conjuction with db. Since this behaviour is what is already in the database, I'm committing this fix for the nested assignment error until I decide what to do about the other issue = val. This now works as it should. I'm still not sure this behaviour is worth the fact that returning mylist from the attribute will actually return a nested structure of PackedLists/Dicts, and these will continue to save to the database whenever they are updated, also if not used in conjuction with db. Since this behaviour is what is already in the database, I'm committing this fix for the nested assignment error until I decide what to do about the other issue..
2012-02-24 23:22:38 +01:00
Griatch
3b88767865
Made fixes to the migrations.
2012-02-15 14:27:26 +01:00
Griatch
a32aebaa0e
OBS- need to run migrations! Refactored attributes to use a slightly different internal storage format for faster access. Also set up caching of all attribute data, so subsequent reads of an attribute will not hit the database anymore, and writes will re-cache.
2012-02-14 23:40:16 +01:00
Griatch
42d502bfc6
OBS - Database schema has changed! If you use south, you need to run manage.py migrate!
...
This adds db indexing on a set of important database fields, all in the interest of optimization.
2012-02-06 13:18:25 +01:00
Griatch
7b2a4e4467
Cleaned up typeclasses, removing a lot of extranenous and ineffective code from the setattr/getattr methods that are called most often.
2012-02-05 21:04:10 +01:00
Griatch
2c4af9076d
Made some commands sensible to the new form of obj.typeclass (it's not a typeclass instance, not a class). Changed the way TypedObject.swap_typeclass works, so as to properly manage failures.
2011-10-03 23:53:23 +02:00
Griatch
bc0195bbaa
Worked with admin interface, cleaning up and adding functionality. There are still some strange behaviour that makes e.g. the presence of inlines to auto-create empty database objects for some strange reason. Inlines are turned off at the moment (so there is no convenient way to add attributes from the admin interface at this time). Creating players now work, but one needs to create all three components (User, Player, Character) in one go and tie them together. The User-admin form was changed to also support multi-word usernames (django default didn't allow spaces).
2011-10-02 22:37:07 +02:00
Griatch
7f2e6dd4fa
Tidying up the admin interface by adding more verbose and helpful names to database fields as well as adding more help texts. The ObjectAdmin still gives tracebacks.
2011-10-02 01:21:03 +02:00
Griatch
23cd9e31b1
Restructured the way typeclasses are loaded. This makes it possible to run at_init() hooks at initiation also for objects without any custom cases for character/players. at_init() hooks are called only when an object is initiated. This means that a room's at_init() hook is only called when someone looks or enters it or a script operates on it, for example, rest of the time these objects are dormant, most efficiently.
2011-10-01 22:00:22 +02:00
Griatch
0a1bcd36c2
Removed FULL_PERSISTENCE setting. It was a "feature" that was added at a time when caching was more inefficient than now. Also the new reload mechanism make FULL_PERSISTENCE=False unfeasable. Use ndb explicitly for non-persistence.
2011-10-01 15:10:21 +02:00
Griatch
2597df0512
Last commit had a typo leading to startup traceback.
2011-09-20 20:26:35 +02:00
Griatch
18669c8700
Fixing so assigning to a nested attribute value doesn't loose data. It is unfortunately not possible to support direct assignment of deeply nested attributes at this time. So if obj.db.mydict = {1:{2:3}}, then obj.db.mydict[1] = 1 will work just fine, but obj.db.mydict[1][2] = 4 will not. For more advanced manipulation such as the latter case, store the dict in a temporary variable and save it back after changes have been done.
2011-09-20 19:59:08 +02:00
Griatch
058f8a9519
Fixed a bug in creating fresh dict-attributes/lists in the new attribute-save system. Resolves issue 191.
2011-09-20 15:55:58 +02:00
Griatch
475361ad28
Added functionality to Attributes to store and update dicts and lists dynamically. One side effect of this is that dicts and tuples need to be stored as custom object types which means that e.g. isintance(obj.db.mylist) == type(list) will return False. In order to do checks like this, use src.utils.utils.inherits_from() instead. The Attribute system now also supports tuples. All other iterables except dicts, lists and tuples are stored and retrieved as lists, same as before.
...
This fixes issue 189.
2011-09-20 12:37:45 +02:00
Griatch
399919a8a9
Database schema has changed. Run migrations if you use South. Started work on tidying up admin interface.
2011-09-14 22:34:10 +02:00
Griatch
f13e8cdf7c
Trunk: Merged griatch-branch. This implements a new reload mechanism - splitting Evennia into two processes: Server and Portal with different tasks. Also cleans and fixes several bugs in script systems as well as introduces i18n (courtesy of raydeejay).
2011-09-03 10:22:19 +00:00
Griatch
2b4e008d18
Scripts and Exits updated. Fixed some deep issues with Scripts that caused object-based scripts to not properly shut down in some situations, as well as spawn multiple instances of themselves. I think this should resolve all "at_repeat doubling" issues reported. Due to optimizations in the typeclass cache loader in a previous update, the Exit cmdsets were not properly loaded (they were loaded at cache time, which now doesn't happen as often). So Exits instead rely on the new "at_cmdset_get" hook called by the cmdhandler. It allows dynamic modification of cmdsets just before they are accessed. Resolves issue173 (I hope). Resolves issue180. Resolves issue 181.
2011-08-11 21:16:35 +00:00
Griatch
ddfd8120bb
Made scripts and typeclassed objects remember db_typeclass_path at all times - a temporarily faulty typeclass will no longer mess up things forever after. Refined and optimized the way typeclasses are cached and loaded, minimizing db hits. The default result when trying to create an object or script with a typeclass that is faulty/not found is now to fail. The previous way, to create an entity anyway using defaults was hard to debug and caused confusion. Resolves issue 175.
2011-08-06 18:15:04 +00:00
Griatch
6cb2b8b745
Working on cleaning some strange behavior when trying to submitting faulty typeclasses to script system. Also fixing bugs here and there.
2011-07-03 21:01:06 +00:00
