migen - Project Hosting on Google Code

Today 12 hours ago r3644 (Added convenience constructor to DefinitionComparisonMessage...) committed by darren.pearce - Added convenience constructor to DefinitionComparisonMessage. Was previously commented out while working on new factored-out approach. Sub-classes now use this. Added convenience constructor to DefinitionComparisonMessage. Was previously commented out while working on new factored-out approach. Sub-classes now use this. 12 hours ago r3643 (DatabasePathClassDatum now allocates child indices automatic...) committed by darren.pearce - DatabasePathClassDatum now allocates child indices automatically rather than requiring it to be specified explicitly in the constructor. Old test rigs for DatabasePath still passed. DatabasePathClassDatum now allocates child indices automatically rather than requiring it to be specified explicitly in the constructor. Old test rigs for DatabasePath still passed. 13 hours ago r3642 (Removed some debug. ) committed by darren.pearce - Removed some debug. Removed some debug. 13 hours ago r3641 (DatabasePath now passes its old test rig. It now uses a Tree...) committed by darren.pearce - DatabasePath now passes its old test rig. It now uses a TreeMap from Integer to the AbstractDatabaseEntity. It also stores the first index (now called a hierarchy index). The Datum class now has a hierarchyIndex (previously 'index') and a childIndex (previously 'key'). This naming is clearer (hopefully!). DatabasePath now passes its old test rig. It now uses a TreeMap from Integer to the AbstractDatabaseEntity. It also stores the first index (now called a hierarchy index). The Datum class now has a hierarchyIndex (previously 'index') and a childIndex (previously 'key'). This naming is clearer (hopefully!). 18 hours ago r3640 (DatabasePath now has new constructor that allows the pre-pen...) committed by darren.pearce - DatabasePath now has new constructor that allows the pre-pending of another definition (database entity) to an existing listing. The sequence of definitions within the paths must all have sequential indices as enforced with the other constructor. This allows a path like <table, column> to be turned into <database, table, column>. To facilitate this new constructor added getIdentifiers() method. Put some comments in about needing to fix getDatabaseName, getTableName, getColumnName, etc. Currently these methods don't work as expected due to the new approach where paths don't always start with a database name. These problems are demonstrated by the TestDatabasePath test-rig in which many of the tests fail at the moment. Fixing this in the next few commits. DatabasePath now has new constructor that allows the pre-pending of another definition (database entity) to an existing listing. The sequence of definitions within the paths must all have sequential indices as enforced with the other constructor. This allows a path like <table, column> to be turned into <database, table, column>. To facilitate this new constructor added getIdentifiers() method. Put some comments in about needing to fix getDatabaseName, getTableName, getColumnName, etc. Currently these methods don't work as expected due to the new approach where paths don't always start with a database name. These problems are demonstrated by the TestDatabasePath test-rig in which many of the tests fail at the moment. Fixing this in the next few commits. 19 hours ago r3639 (Started work on factoring out database-level comparators.) committed by darren.pearce - Started work on factoring out database-level comparators. Started work on factoring out database-level comparators. 19 hours ago r3638 (Removal of redundant import. ) committed by darren.pearce - Removal of redundant import. Removal of redundant import. 19 hours ago r3637 (Removed redundant class TableComparator. The old sub-classes...) committed by darren.pearce - Removed redundant class TableComparator. The old sub-classes now inherit directly from DatabaseEntityComparator. FullTableComparator now extends from new class CompositeDefinitionComparator which is a generic aggregator of comparison messages from a list of 'sub' comparators. Removed redundant class TableComparator. The old sub-classes now inherit directly from DatabaseEntityComparator. FullTableComparator now extends from new class CompositeDefinitionComparator which is a generic aggregator of comparison messages from a list of 'sub' comparators. 19 hours ago r3636 (Removed some redundant imports.) committed by darren.pearce - Removed some redundant imports. Removed some redundant imports. Last 7 days Dec 09, 2009 r3635 (Many changes to the database comparison code. The trigger f...) committed by darren.pearce - Many changes to the database comparison code. The trigger for these changes is that DatabasePath now allows paths like <table, column> rather than requiring that all paths start with a database definition. e.g. <database, table, column> or <database, table>. The empty sequence <> has also been removed. This is due to the generification of the DatabasePath class which now uses the generic parameter to constrain the first argument. In order to handle the empty sequence now, a 'top level' database entity called Catalog has now been defined. All the different constructors in the DatabasePath class have now been removed since the indices defined in the meta-data classes (Datum) are used to ensure that any sequence has a sequence of increasing integers. e.g. <table, column> or <table, primary key>. Note that the refactoring of the DatabaseComparator has not been finished yet. So that everything compiles, the DatabaseDefinitionComparisonReport class is now based on the new generic class DefinitionComparisonMessage. Essentially, the code is in a state of flux between two approaches. I'll finish this off tomorrow morning hopefully. It will be a lot clearer by then. Many changes to the database comparison code. The trigger for these changes is that DatabasePath now allows paths like <table, column> rather than requiring that all paths start with a database definition. e.g. <database, table, column> or <database, table>. The empty sequence <> has also been removed. This is due to the generification of the DatabasePath class which now uses the generic parameter to constrain the first argument. In order to handle the empty sequence now, a 'top level' database entity called Catalog has now been defined. All the different constructors in the DatabasePath class have now been removed since the indices defined in the meta-data classes (Datum) are used to ensure that any sequence has a sequence of increasing integers. e.g. <table, column> or <table, primary key>. Note that the refactoring of the DatabaseComparator has not been finished yet. So that everything compiles, the DatabaseDefinitionComparisonReport class is now based on the new generic class DefinitionComparisonMessage. Essentially, the code is in a state of flux between two approaches. I'll finish this off tomorrow morning hopefully. It will be a lot clearer by then. Dec 09, 2009 run.jnlp (JNLP deployment (r3614b)) file uploaded by mavrikis - Labels: Type-Installer Featured Java-WebStart Labels: Type-Installer Featured Java-WebStart Dec 09, 2009 r3634 (Start of work towards generifying EntityComparison in the ba...) committed by darren.pearce - Start of work towards generifying EntityComparison in the base class and generifying message types. Start of work towards generifying EntityComparison in the base class and generifying message types. Dec 09, 2009 r3633 (Renamed AbstractTableComparator to TableComparator. Is fine ...) committed by darren.pearce - Renamed AbstractTableComparator to TableComparator. Is fine like this since have 'FullTableComparator' rather than going for AbstractTableComparator and TableComparator (for doing the real work). Renamed AbstractTableComparator to TableComparator. Is fine like this since have 'FullTableComparator' rather than going for AbstractTableComparator and TableComparator (for doing the real work). Dec 09, 2009 r3632 (FullTableComparator now chains calls on to the 5 'base' tabl...) committed by darren.pearce - FullTableComparator now chains calls on to the 5 'base' table comparators, collecting all the messages together. FullTableComparator now chains calls on to the 5 'base' table comparators, collecting all the messages together. Dec 09, 2009 r3631 (Added various classes that are moving towards factoring out ...) committed by darren.pearce - Added various classes that are moving towards factoring out the various comparisons that are all currently done with DatabaseDefinitionComparator. Is work in progress at the moment. Will hopefully be finished today/tomorrow. Is part of refamiliarisation/improvement to code clarity. Added various classes that are moving towards factoring out the various comparisons that are all currently done with DatabaseDefinitionComparator. Is work in progress at the moment. Will hopefully be finished today/tomorrow. Is part of refamiliarisation/improvement to code clarity. Dec 09, 2009 r3630 (DatabaseDefinitionComparisonMessage now provides convenience...) committed by darren.pearce - DatabaseDefinitionComparisonMessage now provides convenience constructors. This provision has been moved from DatabaseDefinitionComparisonReport.addMessage() so that the message class can stand alone more easily in some small improvements to this code. DatabaseDefinitionComparator calls to addMessage now changed so use new DatabaseDefinitionComparisonMessage as appropriate. DatabaseDefinitionComparisonMessage now provides convenience constructors. This provision has been moved from DatabaseDefinitionComparisonReport.addMessage() so that the message class can stand alone more easily in some small improvements to this code. DatabaseDefinitionComparator calls to addMessage now changed so use new DatabaseDefinitionComparisonMessage as appropriate. Dec 09, 2009 migensystem.jar (JNLP deployment (3614b)) file uploaded by mavrikis - Labels: Featured Type-Package Labels: Featured Type-Package Dec 09, 2009 migensystem-lib.jar (JNLP deployment (3614b)) file uploaded by mavrikis - Labels: Featured Type-Package Labels: Featured Type-Package Dec 09, 2009 run.jnlp (JNLP deployment (r3614b)) file uploaded by mavrikis - Labels: Featured Type-Installer OpSys-All Labels: Featured Type-Installer OpSys-All Dec 08, 2009 migensystem-1-3629b.zip (migensystem-1-3629b) file uploaded by ToonT...@googlemail.com - Labels: Featured Type-Archive OpSys-All Java-1.X Labels: Featured Type-Archive OpSys-All Java-1.X Dec 08, 2009 r3629 (Added 'Activities World' label to the ecollaborator/activity...) committed by ToonT...@googlemail.com - Added 'Activities World' label to the ecollaborator/activity-document panel. It is in message bundle. Added 'Activities World' label to the ecollaborator/activity-document panel. It is in message bundle. Dec 08, 2009 r3628 (The step amount option 'Random' is now in the message bundle...) committed by ToonT...@googlemail.com - The step amount option 'Random' is now in the message bundle for customisation and localisation. The step amount option 'Random' is now in the message bundle for customisation and localisation. Dec 08, 2009 r3627 (Goal areas in activity documents now use the actual size of ...) committed by ToonT...@googlemail.com - Goal areas in activity documents now use the actual size of the goal area to compute where the subsequent elements should be placed. Eliminates large gap after goal areas. Goal areas in activity documents now use the actual size of the goal area to compute where the subsequent elements should be placed. Eliminates large gap after goal areas. Dec 08, 2009 r3626 (Can now select the step used in animations in the activity d...) committed by ToonT...@googlemail.com - Can now select the step used in animations in the activity document. Choices are Random, 1, 2, 3, or 4. Can now select the step used in animations in the activity document. Choices are Random, 1, 2, 3, or 4. Dec 08, 2009 r3625 (When the animation toolbar is added to an activity document ...) committed by ToonT...@googlemail.com - When the animation toolbar is added to an activity document canvas the canvas component size is reduced rather than the increasing the size of the entire component. Existing task layout no longer affected. When the animation toolbar is added to an activity document canvas the canvas component size is reduced rather than the increasing the size of the entire component. Existing task layout no longer affected. Dec 07, 2009 r3624 (Moved the play (and new step) interface into a toolbar of ac...) committed by ToonT...@googlemail.com - Moved the play (and new step) interface into a toolbar of activity document canvases rather than using the title bar (limited UI options). Moved the play (and new step) interface into a toolbar of activity document canvases rather than using the title bar (limited UI options). Dec 07, 2009 r3623 (Added some documentation to the EntityId class. ) committed by darren.pearce - Added some documentation to the EntityId class. Added some documentation to the EntityId class. Dec 07, 2009 r3622 (Play/Pause menu now replaces play button in all the activity...) committed by ToonT...@googlemail.com - Play/Pause menu now replaces play button in all the activity documents. Play/Pause menu now replaces play button in all the activity documents. Dec 07, 2009 r3621 (The play button is now a menu item in the 'My Document' acti...) committed by ToonT...@googlemail.com - The play button is now a menu item in the 'My Document' activity document. Easily can replace all play buttons with this menu. Menu will have other features soon. The play button is now a menu item in the 'My Document' activity document. Easily can replace all play buttons with this menu. Menu will have other features soon. Dec 06, 2009 r3620 (New system for dealing with attributes in simple cases in pl...) committed by ser...@gmail.com - New system for dealing with attributes in simple cases in place. Because of several ramifications, the new algorithm for the calculation of relation-similarity between simple cases has had to be put in place too. Sorry for the big commit. Attributes are no longer fixed class variables in SimpleCase. Now the names of the attributes of a simple case and their initial values are set at creation time. This allows the same class SimpleCase to be used with other microworlds apart from eXpresser (e.g. ShapeBuilder). Identity attributes are in place. Identity distances are calculated in SimpleCaseSimilarityMeasurer, but are not integrated yet in CompositeCaseSimilarityMeasurer. This will be done after the relation similarity is fixed (see bellow). The new relation similarity has been implemented. - In the former scheme, shapes were related to shapes (via the AttributeExpressions, aka tied numbers). This presented several problems of directionality (i.e. if A.iterations, B.iterations and C.moveDown use the same number, do the pairs A-B, B-C, and C-A have the same relation similarity?) and cardinality (i.e. if A.iterations, A.moveDown, B.iterations and C.iterations use the same number, do the pairs A-B, and B-C have the same relation similarity?). - The new scheme links attributes with attributes. When two simple cases are compared for relations similarity, a list is calculated for each attribute to see which other attributes (in the same or other shapes) are connected to it via AttributeExpression). For each case, a Relation (new class in this commit) is created. Relations are symmetrical, therefore if we compare A (with some A') in the first example above, only two relations would be created for A (i.e. A.iterations-B.iterations, A.iterations-C.moveDown); but if we compare A (with some A') in the second example, five relations would be created (i.e. A.iterations-A.moveDown, A.iterations-B.iterations, A.iterations-C.iterations, A.moveDown-B.iterations, A.moveDown-C.iterations). This relations are compared to find the similarity between A and A' using the Jacquard index (i.e. rels(A )-intersection-rels(A') / rels(A)-union-rel(A')). This new calculation of the relations similarity does not work yet, because expressions are not taking into account the links of subexpressions (i.e. it is not implemented yet that "5 + 4" is linked to the attributes where that "4" is used, not just to the attributes where "5 + 4" is used). This did not affect the former calculation of relations, but it affects the current one. This will be implemented in the next commit (so this commit does not work perfectly, but at least it compiles). The current scheme only works for constructions in which only expressions at the highest level are shared. Another important change is the clear distinction between distances and shapes when calculating similarities (especially SimpleCaseSimilarityMeasurer); and the fact that distances are now weighted (both the numerical distance and the identity distance). Of course, in between all these major changes there is plenty of minor changes: new methods, improved javadoc in some old methods, a couple of new classes (RelationEnd and Relation), etc. New system for dealing with attributes in simple cases in place. Because of several ramifications, the new algorithm for the calculation of relation-similarity between simple cases has had to be put in place too. Sorry for the big commit. Attributes are no longer fixed class variables in SimpleCase. Now the names of the attributes of a simple case and their initial values are set at creation time. This allows the same class SimpleCase to be used with other microworlds apart from eXpresser (e.g. ShapeBuilder). Identity attributes are in place. Identity distances are calculated in SimpleCaseSimilarityMeasurer, but are not integrated yet in CompositeCaseSimilarityMeasurer. This will be done after the relation similarity is fixed (see bellow). The new relation similarity has been implemented. - In the former scheme, shapes were related to shapes (via the AttributeExpressions, aka tied numbers). This presented several problems of directionality (i.e. if A.iterations, B.iterations and C.moveDown use the same number, do the pairs A-B, B-C, and C-A have the same relation similarity?) and cardinality (i.e. if A.iterations, A.moveDown, B.iterations and C.iterations use the same number, do the pairs A-B, and B-C have the same relation similarity?). - The new scheme links attributes with attributes. When two simple cases are compared for relations similarity, a list is calculated for each attribute to see which other attributes (in the same or other shapes) are connected to it via AttributeExpression). For each case, a Relation (new class in this commit) is created. Relations are symmetrical, therefore if we compare A (with some A') in the first example above, only two relations would be created for A (i.e. A.iterations-B.iterations, A.iterations-C.moveDown); but if we compare A (with some A') in the second example, five relations would be created (i.e. A.iterations-A.moveDown, A.iterations-B.iterations, A.iterations-C.iterations, A.moveDown-B.iterations, A.moveDown-C.iterations). This relations are compared to find the similarity between A and A' using the Jacquard index (i.e. rels(A )-intersection-rels(A') / rels(A)-union-rel(A')). This new calculation of the relations similarity does not work yet, because expressions are not taking into account the links of subexpressions (i.e. it is not implemented yet that "5 + 4" is linked to the attributes where that "4" is used, not just to the attributes where "5 + 4" is used). This did not affect the former calculation of relations, but it affects the current one. This will be implemented in the next commit (so this commit does not work perfectly, but at least it compiles). The current scheme only works for constructions in which only expressions at the highest level are shared. Another important change is the clear distinction between distances and shapes when calculating similarities (especially SimpleCaseSimilarityMeasurer); and the fact that distances are now weighted (both the numerical distance and the identity distance). Of course, in between all these major changes there is plenty of minor changes: new methods, improved javadoc in some old methods, a couple of new classes (RelationEnd and Relation), etc. Last 30 days Dec 04, 2009 r3619 (Minor changes. - Improved Javadoc. - Some new methods that ...) committed by ser...@gmail.com - Minor changes. - Improved Javadoc. - Some new methods that are needed for the new generic implementation of simple cases' attributes. Committed as a last-point-before-failure. Changes needed now for the new scheme are going to break the implementation; nothing will be committed until the changes have been finished and the new implementation is working as the current one. Minor changes. - Improved Javadoc. - Some new methods that are needed for the new generic implementation of simple cases' attributes. Committed as a last-point-before-failure. Changes needed now for the new scheme are going to break the implementation; nothing will be committed until the changes have been finished and the new implementation is working as the current one. Dec 02, 2009 r3618 (The result from the 'How many?' menu item for expressions no...) committed by ToonT...@googlemail.com - The result from the 'How many?' menu item for expressions now updates when the embedded tied numbers or the expression is changed. These kinds of numbers have a different background colour. Also the distance between check boxes in goal boxes is reduced. The result from the 'How many?' menu item for expressions now updates when the embedded tied numbers or the expression is changed. These kinds of numbers have a different background colour. Also the distance between check boxes in goal boxes is reduced. Dec 02, 2009 r3617 ('How many?' menu item for expressions now displays a number ...) committed by ToonT...@googlemail.com - 'How many?' menu item for expressions now displays a number that is different in the slave panel (General World). It doesn't yet consistently update when the expression is changed. 'How many?' menu item for expressions now displays a number that is different in the slave panel (General World). It doesn't yet consistently update when the expression is changed. Dec 02, 2009 r3616 (When a tab is selected in a property list then the slave mir...) committed by ToonT...@googlemail.com - When a tab is selected in a property list then the slave mirror selects the corresponding tab. When a tab is selected in a property list then the slave mirror selects the corresponding tab. Dec 02, 2009 r3615 (Menu item label for subtraction now distinguishes between dr...) committed by ToonT...@googlemail.com - Menu item label for subtraction now distinguishes between dropping on the left or right. Property lists now have separate tabs for construction attributes and translation attributes. Slave panel (General World) no longer has a mirror of the pattern creation wizard. (Multiple changes since could not commit earlier due to server being down.) Menu item label for subtraction now distinguishes between dropping on the left or right. Property lists now have separate tabs for construction attributes and translation attributes. Slave panel (General World) no longer has a mirror of the pattern creation wizard. (Multiple changes since could not commit earlier due to server being down.) Nov 30, 2009 migensystem-1-3614b.zip (migensystem-1-3614b (first implementation of model undo and ...) file uploaded by ToonT...@googlemail.com - Labels: Featured Type-Archive Java-1.X OpSys-All Labels: Featured Type-Archive Java-1.X OpSys-All Nov 30, 2009 r3614 (Renamed menu item from 'Create Snapshot' to 'Copy'.) committed by ToonT...@googlemail.com - Renamed menu item from 'Create Snapshot' to 'Copy'. Renamed menu item from 'Create Snapshot' to 'Copy'. Nov 30, 2009 r3613 (Changed 'Rule' label to 'World Colouring Rule' and removed t...) committed by ToonT...@googlemail.com - Changed 'Rule' label to 'World Colouring Rule' and removed the 'How many?' labels from rules. Changed 'Rule' label to 'World Colouring Rule' and removed the 'How many?' labels from rules. Nov 30, 2009 issue 666 (load pre-saved or programmatically construct expressions in ...) commented on by ToonT...@googlemail.com - Or should they have blocked 616? ;-) In May 2005, it was reported that scholars at Oxford University using advanced imaging techniques[19] had been able to read previously illegible portions of the earliest known record of the Book of Revelation (a 1,700 year old papyrus), from the Oxyrhynchus site, Papyrus 115 or P115, dating one century after Irenaeus. The fragment gives the Number of the Beast as 616 (chi, iota, sigma), rather than the majority text 666 (chi, xi, sigma). http://en.wikipedia.org/wiki/Number_of_the_Beast Or should they have blocked 616? ;-) In May 2005, it was reported that scholars at Oxford University using advanced imaging techniques[19] had been able to read previously illegible portions of the earliest known record of the Book of Revelation (a 1,700 year old papyrus), from the Oxyrhynchus site, Papyrus 115 or P115, dating one century after Irenaeus. The fragment gives the Number of the Beast as 616 (chi, iota, sigma), rather than the majority text 666 (chi, xi, sigma). http://en.wikipedia.org/wiki/Number_of_the_Beast Nov 30, 2009 r3612 (Introduced a new configuration option undo-enabled (default ...) committed by ToonT...@googlemail.com - Introduced a new configuration option undo-enabled (default is true) which enables the undo and redo menu items (under Edit). Control-z is a short cut for undo. This implements undo of changes to the model -- not changes to the canvas (e.g. dropping expressions on the canvas). Note that this implementation has the advantage (and disadvantage) that it saves XML files for every state of the model. Could in principle be used to replay a session. Introduced a new configuration option undo-enabled (default is true) which enables the undo and redo menu items (under Edit). Control-z is a short cut for undo. This implements undo of changes to the model -- not changes to the canvas (e.g. dropping expressions on the canvas). Note that this implementation has the advantage (and disadvantage) that it saves XML files for every state of the model. Could in principle be used to replay a session. Nov 30, 2009 issue 666 (load pre-saved or programmatically construct expressions in ...) commented on by ser...@gmail.com - And there I was wondering whether google would allow issues with this number... And there I was wondering whether google would allow issues with this number... Nov 29, 2009 issue 669 (sequencing tasks within an activity ) changed by ToonT...@googlemail.com - This issue was closed by revision r3611. Status: Fixed This issue was closed by revision r3611. Status: Fixed Nov 29, 2009 r3611 (Done button is a goal area in an activity document is now on...) committed by ToonT...@googlemail.com - Done button is a goal area in an activity document is now only enabled when all the check boxes are ticked. Resolves Issue 669 . Done button is a goal area in an activity document is now only enabled when all the check boxes are ticked. Resolves Issue 669 . Nov 29, 2009 r3610 (task-activity-document-N now takes an additional optional ar...) committed by ToonT...@googlemail.com - task-activity-document-N now takes an additional optional argument which is the name of the next activity that should be started when the DONE button is pressed. Default configuration updated so that finishing Train Track starts Footpath which when finished starts Garden. Does most of Issue 669 . task-activity-document-N now takes an additional optional argument which is the name of the next activity that should be started when the DONE button is pressed. Default configuration updated so that finishing Train Track starts Footpath which when finished starts Garden. Does most of Issue 669 . Nov 29, 2009 r3609 (Abstracted goal areas in activity documents into their own c...) committed by ToonT...@googlemail.com - Abstracted goal areas in activity documents into their own class. No change in functionality yet. Also added the DONE button label to the Message Bundle. Abstracted goal areas in activity documents into their own class. No change in functionality yet. Also added the DONE button label to the Message Bundle. Nov 29, 2009 issue 675 (each activity should open a new page ) changed by ToonT...@googlemail.com - This issue was closed by revision r3608. Status: Fixed This issue was closed by revision r3608. Status: Fixed Nov 29, 2009 r3608 (A new tab page is opened every time the user selects an acti...) committed by ToonT...@googlemail.com - A new tab page is opened every time the user selects an activity. Resolves Issue 675 . A new tab page is opened every time the user selects an activity. Resolves Issue 675 . Nov 29, 2009 issue 675 (each activity should open a new page ) commented on by ToonT...@googlemail.com - I wonder if it should open a new page only when the activity loads a model in My World. I wonder if it should open a new page only when the activity loads a model in My World. Nov 29, 2009 issue 161 (Should all ExpressedObjects have a location) commented on by darren.pearce - Yes it is very old! Is a relic of when the software was a lot more generic. Now we are permanently in 'blocks world', we don't really need to worry about this since it is true for all our objects. Yes it is very old! Is a relic of when the software was a lot more generic. Now we are permanently in 'blocks world', we don't really need to worry about this since it is true for all our objects. Nov 29, 2009 r3607 (New approach to attributes of Simple Cases, more general and...) committed by ser...@gmail.com - New approach to attributes of Simple Cases, more general and scalable. Attributes of a simple case can be numeric, identity, or relation. Numeric attributes are integers, in some cases defined by expressions. They are added. Identity attributes can be anything. They are checked for equality. All simple cases must have the same attributes. New approach to attributes of Simple Cases, more general and scalable. Attributes of a simple case can be numeric, identity, or relation. Numeric attributes are integers, in some cases defined by expressions. They are added. Identity attributes can be anything. They are checked for equality. All simple cases must have the same attributes.

r3620 (New system for dealing with attributes in simple cases in pl...) committed by ser...@gmail.com - New system for dealing with attributes in simple cases in place. Because of several ramifications, the new algorithm for the calculation of relation-similarity between simple cases has had to be put in place too. Sorry for the big commit. Attributes are no longer fixed class variables in SimpleCase. Now the names of the attributes of a simple case and their initial values are set at creation time. This allows the same class SimpleCase to be used with other microworlds apart from eXpresser (e.g. ShapeBuilder). Identity attributes are in place. Identity distances are calculated in SimpleCaseSimilarityMeasurer, but are not integrated yet in CompositeCaseSimilarityMeasurer. This will be done after the relation similarity is fixed (see bellow). The new relation similarity has been implemented. - In the former scheme, shapes were related to shapes (via the AttributeExpressions, aka tied numbers). This presented several problems of directionality (i.e. if A.iterations, B.iterations and C.moveDown use the same number, do the pairs A-B, B-C, and C-A have the same relation similarity?) and cardinality (i.e. if A.iterations, A.moveDown, B.iterations and C.iterations use the same number, do the pairs A-B, and B-C have the same relation similarity?). - The new scheme links attributes with attributes. When two simple cases are compared for relations similarity, a list is calculated for each attribute to see which other attributes (in the same or other shapes) are connected to it via AttributeExpression). For each case, a Relation (new class in this commit) is created. Relations are symmetrical, therefore if we compare A (with some A') in the first example above, only two relations would be created for A (i.e. A.iterations-B.iterations, A.iterations-C.moveDown); but if we compare A (with some A') in the second example, five relations would be created (i.e. A.iterations-A.moveDown, A.iterations-B.iterations, A.iterations-C.iterations, A.moveDown-B.iterations, A.moveDown-C.iterations). This relations are compared to find the similarity between A and A' using the Jacquard index (i.e. rels(A )-intersection-rels(A') / rels(A)-union-rel(A')). This new calculation of the relations similarity does not work yet, because expressions are not taking into account the links of subexpressions (i.e. it is not implemented yet that "5 + 4" is linked to the attributes where that "4" is used, not just to the attributes where "5 + 4" is used). This did not affect the former calculation of relations, but it affects the current one. This will be implemented in the next commit (so this commit does not work perfectly, but at least it compiles). The current scheme only works for constructions in which only expressions at the highest level are shared. Another important change is the clear distinction between distances and shapes when calculating similarities (especially SimpleCaseSimilarityMeasurer); and the fact that distances are now weighted (both the numerical distance and the identity distance). Of course, in between all these major changes there is plenty of minor changes: new methods, improved javadoc in some old methods, a couple of new classes (RelationEnd and Relation), etc.

New system for dealing with attributes in simple cases in place. Because of several ramifications, the new algorithm for the calculation of relation-similarity between simple cases has had to be put in place too. Sorry for the big commit. Attributes are no longer fixed class variables in SimpleCase. Now the names of the attributes of a simple case and their initial values are set at creation time. This allows the same class SimpleCase to be used with other microworlds apart from eXpresser (e.g. ShapeBuilder). Identity attributes are in place. Identity distances are calculated in SimpleCaseSimilarityMeasurer, but are not integrated yet in CompositeCaseSimilarityMeasurer. This will be done after the relation similarity is fixed (see bellow). The new relation similarity has been implemented. - In the former scheme, shapes were related to shapes (via the AttributeExpressions, aka tied numbers). This presented several problems of directionality (i.e. if A.iterations, B.iterations and C.moveDown use the same number, do the pairs A-B, B-C, and C-A have the same relation similarity?) and cardinality (i.e. if A.iterations, A.moveDown, B.iterations and C.iterations use the same number, do the pairs A-B, and B-C have the same relation similarity?). - The new scheme links attributes with attributes. When two simple cases are compared for relations similarity, a list is calculated for each attribute to see which other attributes (in the same or other shapes) are connected to it via AttributeExpression). For each case, a Relation (new class in this commit) is created. Relations are symmetrical, therefore if we compare A (with some A') in the first example above, only two relations would be created for A (i.e. A.iterations-B.iterations, A.iterations-C.moveDown); but if we compare A (with some A') in the second example, five relations would be created (i.e. A.iterations-A.moveDown, A.iterations-B.iterations, A.iterations-C.iterations, A.moveDown-B.iterations, A.moveDown-C.iterations). This relations are compared to find the similarity between A and A' using the Jacquard index (i.e. rels(A )-intersection-rels(A') / rels(A)-union-rel(A')). This new calculation of the relations similarity does not work yet, because expressions are not taking into account the links of subexpressions (i.e. it is not implemented yet that "5 + 4" is linked to the attributes where that "4" is used, not just to the attributes where "5 + 4" is used). This did not affect the former calculation of relations, but it affects the current one. This will be implemented in the next commit (so this commit does not work perfectly, but at least it compiles). The current scheme only works for constructions in which only expressions at the highest level are shared. Another important change is the clear distinction between distances and shapes when calculating similarities (especially SimpleCaseSimilarityMeasurer); and the fact that distances are now weighted (both the numerical distance and the identity distance). Of course, in between all these major changes there is plenty of minor changes: new methods, improved javadoc in some old methods, a couple of new classes (RelationEnd and Relation), etc.

Today

Last 7 days

Last 30 days