Dodeca PBCS/EPM Cloud Connector

Fixed Zoom In expanding only one member of a selection. Selecting several members — by dragging a range over them or by selecting them separately — and zooming in now expands every one of them, as it does against Essbase. Previously only one was expanded and the rest were left as they were, so a range had to be zoomed one member at a time. Members that were not selected keep their places, and cells in the selection that hold no member, such as the data beside a dragged range, are ignored. #3924

Fixed Zoom Out acting only on the first of several selected members. Selecting several members, by dragging a range over them or by selecting them separately, and zooming out now collapses every one of them into its parent, as it does against Essbase. Previously only the first was zoomed out.

Zoom In and Zoom Out now act on members selected on the rows and the columns at once, and on members of two different row dimensions in the same row, as they do against Essbase. Previously only the members in the part of the grid where the selection began were zoomed, and only the first member in a row.

Fixed Zoom In on a member that repeats under each of several outer members, such as an account under each entity, interleaving the results. Each occurrence now expands in place under its own outer member, so the children appear grouped by entity, as they do against Essbase. Previously, occurrences in adjacent rows were treated as one block, and the children came out interleaved: the first child under every entity, then the second under every entity, and so on. #4252

Fixed Zoom Out collapsing the wrong rows and columns. Zooming out a member now collapses it, together with its siblings and anything under them, into their parent wherever they appear, each in place under its own outer members, on the rows and on the columns alike, as it does against Essbase. Previously, Zoom Out could take away a member that was not part of the group, such as the next member below it, its cousins beside it, or the same members under the next entity. It could also drop all but the first member of a dimension nested inside the group. On the columns, only the selected member changed, to its parent, leaving its siblings beside it. #4252

Zoom In to all levels or to the bottom level stops at a shared member, as it does against Essbase, rather than going on into the descendants of the member it shares. Earlier releases could unroll the whole of a shared member’s base hierarchy beneath it. Zooming in on a shared member itself expands the member it shares, as Dodeca does against Essbase. The EPM Cloud application settings "Allow Drill Down on Shared Members in Ad Hoc" and "Drill to All Levels in Base" are not applied. #4042

The Remove Unselected Groups option now applies against EPM Cloud. With it on, zooming in keeps the members that were selected and what they expand into, and removes the other members of that dimension, as it does against Essbase. With Within Selected Group on as well, only the group the selection is in is affected. Previously the option was ignored.

Fixed pivoting a column member when another dimension is also on the columns. The member’s dimension now moves to the front of the rows and the other column dimension stays on the columns, as it does against Essbase. Previously the two column dimensions swapped places and nothing reached the rows. Pivoting a row member when there is more than one dimension on the rows likewise moves its dimension to the front of the columns, as Essbase does. #4384, #4120

Fixed pivoting the outermost column dimension of a grid with no POV failing with an EPM Cloud error that the dimension or its members were missing. The dimension now moves to the front of the rows, as it does against Essbase. Previously it was moved into a POV the grid did not have, and so was dropped from the retrieve altogether. #4128

Fixed pivoting a row member when another dimension is also on the rows. Its dimension now reaches the columns carrying every one of its members, and the rows it leaves behind close up, so each remaining row appears once, as it does against Essbase. Previously only the member that was clicked reached the columns, and each remaining row was repeated once for every member that should have gone with it. Pivoting a column member to the rows closes up the columns in the same way. #4383

Fixed the top column member moving one cell to the left on a retrieve when two or more dimensions, attribute dimensions included, are on the rows and every column header is stacked over a single data column. A POV row now stays where it was typed, as it does against Essbase. Previously it was always written back from the second column, under the row labels once there was more than one row dimension. When a pivot or zoom adds a row dimension or takes one away, the POV row moves along with the row dimensions, as Essbase moves it. #4303

Fixed member labels appearing on the first two rows instead of only the first after pivoting the only member of the POV onto the rows with Repeat Member Labels turned off. The labels of the outer row members now appear only where they change, as they do against Essbase. #4165

Pivoting the only dimension on the columns now moves it to the rows, as it does against Essbase, rather than being refused with "Cannot pivot last row in top region". The dimension goes to the front of the rows carrying all of its members, and the grid is left with nothing on its columns: a single, unlabelled data column, which can be retrieved, zoomed and pivoted like any other grid. The dimension stays on the columns only when nothing is in the POV, because EPM Cloud needs a POV member in place of the missing column to retrieve such a grid. #4380

Fixed a pivot that names where the dimension should go, such as one from a workbook script given a source cell and a target cell, ignoring the target. The dimension now moves to the place the target cell names — the POV, the rows or the columns, and the position within them — as it does against Essbase. Previously the target was ignored and the dimension moved as though it had been pivoted from the toolbar, so a script could not choose where it landed. Pivots from the toolbar are unchanged.

Fixed a Dimensional Query Expression Pov that names more than one member of a dimension, such as several months, using only the last of them. Pov now keeps every member it is given, and Data returns the members that have data in any of them, as a grid of those months shows with missing rows suppressed. Previously a member with data in earlier months but none in the last month listed was left out, and adding a month with no data to the end of the list could make the whole result empty. A Pov naming one member of each dimension is unchanged. #4354

Blank rows and columns now stay between the rows and columns they separated through Zoom In, Zoom Out, Keep Only, pivots and suppressed rows, as they do against Essbase. A blank stays immediately before the first remaining row that came after it, and goes away when nothing after it remains. Previously a blank was put back at its old row number. So zooming in on a member above a blank split the new family around it, and zooming out above one left it at the bottom of the grid. A row of unrecognized members is never dropped this way; with nothing after it, it stays at the end.

Fixed a retrieve returning a grid one column wider than it was sent, with an empty column on the right, when a blank column sits between its data columns and the POV row is the widest part of the grid. The blank column now stays where it was and the grid keeps its width.

A grid that names a dimension in more than one place, such as an entity in the POV and another down the rows, is now refused with an error that names the dimension and both places, as Essbase refuses such a grid. Previously the retrieve came back with no data rows and no error.

Fixed an error whose message contains &, < or >, such as an unknown member named R&D, reaching Dodeca as an unreadable response instead of the error itself. The message now arrives intact.

Fixed a Dimensional Query Expression Data of a cross join returning only its first column. Each tuple is now checked at its own intersection, and the tuples with data come back whole, with every column. Previously only the members of the first dimension came back, and they were checked against the top of the other dimensions rather than the members they were paired with. #4435

Fixed the dimension numbers in Data Audit Log records counting from zero while the cube information the connector gives Dodeca counts from one, as Essbase does. Dodeca names the dimensions in a logged change by looking their numbers up in the cube information, so each dimension would have been named after the one before it. Records now number dimensions from one. Records written by earlier releases still count from zero. #4002

Reduced the number of EPM Cloud requests made to resolve the members on a grid. Members a grid has already resolved are remembered for the life of that grid, so repeating a retrieve no longer resolves every row and column header again.

Fixed member cells never being marked as dimension members. A cell holding a dimension’s own member — Entity in the Entity dimension, say — is meant to carry a flag saying so, and never did, for any grid. The check compared the member’s name against the cell’s value after that value had been quoted for transmission, so it could not match. Clients that distinguish a dimension member from an ordinary one now receive what they expect.

Fixed attribute dimensions being counted among the dimensions a grid has to place. A grid may carry an attribute dimension and zoom or pivot it like any other, but it is complete without one, so attribute dimensions no longer count towards the minimum a layout requires. Previously a cube whose standard dimensions numbered fewer than two, but which had attribute dimensions, was accepted and then could not produce a grid.

The connector now names Tomcat as its servlet container. It had declared the Jetty starter and shipped the Jetty libraries, but Tomcat is what it actually started: the bundled RetroEss dependency put Tomcat’s classes on the classpath, and Spring Boot selects a container by what it finds there, choosing Tomcat before Jetty. Behavior is unchanged, since Tomcat was already the container in use, but the Jetty libraries are no longer shipped and any Jetty-specific configuration would previously have been ignored.

Reduced what the connector bundles by removing unreachable code and two unused dependencies from RetroEss.

Updated to PBJ client library version 3.0.4.

Known issue: when a row or column of a grid holds a member EPM Cloud does not recognize, the Unknown Member Explorer can list the other members in that row or column as unknown too, even though they are valid. #3975

Removed an unused ad hoc grid engine from the bundled RetroEss dependency. RetroEss carried a second grid interpreter and its own set of ad hoc operations, written for the HFM connector and never called by this connector, which performs all grid work through JayGrid. Removing it reduces what the connector ships without changing its behavior.

Renamed the connector’s documentation pages to descriptive slugs and pointed the distribution assembly at the renamed PDF.

Added optional parallel execution for large EPM Cloud retrieves that exceed maxCellsPerRetrieve and must be split into multiple REST requests. The maxConcurrentRetrieveRequests connection parameter controls per-retrieve concurrency, while a shared, bounded connector executor limits aggregate work. The default remains one concurrent request for compatibility. This also fixes maxCellsPerRetrieve connection values not being propagated to the retrieve operation. #4212

Fixed Zoom In workbook-script operations failing to expand a selected left-axis member in fully stacked ad hoc grids without a POV. Zoom targeting now uses the actual POV height, with regression coverage for the no-POV layout. #3883

Fixed a regression where enabling Show Empty Grid Error did not report an error when suppress-missing or suppress-zero removed every data row. Empty ad hoc results now raise an error before the existing worksheet grid is overwritten, while explicitly permitted empty grids and dataless navigation remain unchanged. #4127

Fixed Keep Only and Remove Only operations selecting the wrong left-axis row in fully stacked ad hoc grids without a POV. Left-axis selections now account for the grid’s actual POV height instead of assuming a single POV row, which also corrects targeting for vertically stacked POVs. #4129

Added an opt-in, bearer-token-protected POST /admin/cache/clear endpoint for scripts and administrative tooling to clear all EPM Connector member caches without restarting the service. The endpoint is disabled by default. #4261

Changed the clearCache=true connection parameter to clear member metadata only for the configured EPM Cloud server, application, and plan. Previously, using the parameter also cleared unrelated plans served by the connector. #4261

Added support for non-Default EPM Cloud alias tables. Administrators can declare additional known tables with the semicolon-delimited aliasTables connection parameter, and views honor the selected EssProperties.AliasTable. The configured names are trusted without making validation requests because Oracle EPM Cloud does not provide a supported REST API for enumerating alias tables. The parameter may be used without an explicit dimensions list, in which case dimensions are discovered automatically. #4568

Fixed an issue where testing or using a Dimensional Query Expression with UseAliases enabled returned member names instead of aliases. DQEs without an explicit output format now return aliases from the selected alias table, falling back to the member name when an alias is unavailable, while explicit MBRNAMES and ALTNAMES format selections remain unchanged. #4035

Fixed an audit-response compatibility issue when submitting values into previously empty EPM Cloud cells. EPM Cloud exports missing values as blank strings; the connector omits those missing prior values from the audit XML rather than emitting empty numeric elements, allowing Dodeca to deserialize the response and write Data Audit Log records. The underlying correction was introduced in version 1.7.16, and version 2.0.3 adds explicit regression coverage for this scenario. #4006

Fixed an upstream PBJ library issue where member hierarchies could include members assigned only to other plan types. #4495

Fixed erratic row indentation when a view’s Indentation property is set to Totals. Totals indentation is derived from member levels, which could be inflated by the member hierarchy issue above; filtering hierarchies by the selected plan type corrects the indentation, and regression coverage verifies each indentation style. #4079

Fixed inclusive children and descendants in the member selection dialog to return parents after their children, matching Essbase Report Writer ordering. #4253

Fixed issue when trying to run a calc/business rule that doesn’t use any parameters/runtime substitution variables. #4322

Improved the error message when an EssbaseRunCalc ServerBased operation requests Dodeca token replacement. EPM Cloud can launch deployed business rules by name and pass runtime prompts, but its REST API cannot retrieve a business rule as calc-script text for token replacement. #4344

Fixed an issue that prevented Dimensional Query Expressions containing line breaks from being evaluated. #4099

Added documentation for the DQE Hierarchy function and its NATURAL and POST ordering options. #4309

Documented that DQE expressions are created and used as Essbase Member Query scripts in Dodeca. #4116

Fixed issue when using "By Level Name" or "By Generation Name" in a member selection dialog that would cause an exception. #4224, #4226, #4289

Fixed several issues related to using the Find functionality in selector lists, with and without the "Shared Members" option being checked. These fixes include errors such as "Member Entity/Total Entity/TD/840 (or other member names) was not found in the tree", issues when using the SharedMembersCheckboxVisible option in a selector list, and others. #4012, #4029, #4207

The DQE Leaves function will now return members in outline order by default, to more closely mirror traditional Essbase behavior, and to obviate the need to wrap it in a Hierarchy function. #4269

Fixed an issue where an erroneous exception would be reported instead of the underlying cause (connector would report a NoClassDefFoundError or similar message instead of a true error message). #4267

This version updates to newer version of the Spring Framework, and also requires Java 17 or higher as a minimum version.

Updated to PBJ client version 3.0.0, which includes a fix to avoid a "cannot retry due to server authentication, in streaming mode" exception that could be thrown in certain circumstances. #4178

The connector will now return a better error message instead of "Error retrieving grid: Members Hash Empty". This error would be returned when all row members were invalid, and the underlying request to the EPM Cloud REST API would ask for a grid with no members in the row or column orientation. #4181

A better error message will be returned if there is an invalid member in the POV. Previously a com.appliedolap.jaygrid.Grid.getCell(int, int) is null error message would be returned. The connector does not support invalid members in the POV. #4193

Setting the Essbase log level in Dodeca to DEBUG or higher will now cause the PBJ payloads to the EPM Cloud REST API to be logged, which can assist troubleshooting. #4196, #4211

The connector will now return a list of calculation scripts (business rules) if requested by a DAI/DSMS client. This prevents some exceptions when exploring objects but support for accessing their contents does not exist. #4200

Fixed an issue when testing an DQE script (as an MDX script) that would cause an exception when 'return members as list' was checked. #4227

Better error messages are returned if a Dodeca client makes a call to the connector that is not supported, instead of providing a generic exception message. #4228

Fixed issue where outline members that are marked as invalid (due to not existing in the outline but at attempt to retrieve them was made) could not be ever marked as valid, unless the connector was restarted. By default, invalid members will be reevaluated for invalidity after 15 minutes. This setting can be controlled with the connector.cache.refresh-interval setting in the configuration. #4229

Fixed issue where the DQE Distinct() function would return duplicates. #4236

Fixed issue where retrieves from FCCS would not contain any data (typically when doing a retrieve with a reporting currency in the grid layout). The underlying issue is due to duplicate member/alias names. The connector will no longer attempt to cache alias names that would displace a cache entry for an actual member name.

Added maxCellsPerRetrieve connection parameter (e.g., database setting might be Basic?dimensions=Scenario;Period&maxCellsPerRetrieve=200000) to allow explicit configuration of how many cells are used in a single retrieve. The EPM Cloud REST API has an upper limit of 500,000 cells, including member cells. The default if not explicitly configured is 400,000 (that allows for comfortably being under the limit). Administrators may want to set this lower so that retrieves that are broken up into multiple pages are smaller, which may yield faster overall retrieve times. #4105

Fixed issue using find member dialog from the Dodeca Add-In for Excel. #4154

Added a CrossJoin function to DQE that can help generate member permutations. #4156

Fixed an issue with the Data DQE where wasn’t building a dimension list properly, causing an exception to be thrown. #4074

Add additional logging to DQE methods (debug level).

Added option to Hierarchy DQE function to specify member ordering. Options are NATURAL and POST, e.g. Hierarchy(Descendants(Year), POST). The default option is NATURAL, such as Hierarchy(Descendants(Year)). #4075

Added POV DQE function to generate a member set for a POV in a cube. By default, a POV at the "top of the house" (e.g. top-level members for each dimension, such as (Account, Time, Entity, Scenario, Version)) is generated, but zero or more optional parameters will narrow it, such as POV('Jan', 'FY2025') which will be a fully qualified POV for the given members, plus top-of-the-house for all unspecified dimensions. #4074

Added Data DQE to filter a member set to only include intersections where data exists. This function can be used for a single POV (cell) or for a POV along with a list of members for a single dimension. For example data(pov(), member('Jan', 'Feb', 'Mar')) will generate a top-of-the-house POV, then check for data for three cells that are the permutations of that POV plus the given time periods. #4074

Fixed an issue where a selector list would not include original member when it used a <DESCENDANTSOF query. #4059

Fixed an issue where retrieve ranges with additional blank rows would improperly add additional blank rows on a zoom-in operation. #3996

Fixed issue when zooming/operating on grid with member names and aliases. #4057

Added experimental Hierarchy function to DQEs to sort a member set according to member order in the dimension. Note: performance of this method is not extensively tested, large dimensions may incur a lag in processing. #4039

Fixed several issues with data update/submit if blank/missing values were used, blank columns/rows, and repeating member names. #4058

Add support for sending blanks and zero as missing values when submitting/updating data. #4063

Retrieves are now more robust to try to avoid an index out of bounds error. #4050

Fixed issue where zoom-in operations would suppress shared members. #4051

Fixed issue where Find in Member Selection dialog would only find one copy of a member. #4048

Fixed issue where columns would be arranged incorrectly when retrieving a grid containing members that the user does not have access to. Note: cells that a user does not have access to appear as #Missing, as there is presently no way for the connector to differentiate between #NoAccess and #Missing. #4046

Finding members (using a selector) has been changed to default to case-insensitive search (underlying PBJ library updated). #4027

Member search (such as performed in a selector list) now searches aliases in addition to member names. #4028

Fixed issue where member selection on a "top"/col dimension would throw exception. #4009

Fixed issue where extra members would be added to a retrieve/zoom when suppress missing is turned on. #4025

Fixed issue when navigate without data option is enabled. #4026

Fixed issue when using label/invalid rows and suppress missing is turned on causing additional rows to get cleared

Fixed issue when using suppress missing when use both member/alias is turned on

Fix support for member selection dialog. #3993

Provide support for basic member selection rules in member selection dialog (children, children including member, descendants, descendants including member). The "subset" rule does not work.

Add support for Preview button in member selection dialog

Support for dynamic logging level (Log Level for Current Session option in client)

Using Find inside a selector list now works (to search for members in by name)

Add additional logging to zoom in (enable debug or higher to generate)

Fixed issue preventing with data audit logging updates not being sent to Dodeca server

Fix issue where incomplete rows (e.g. a row meant to be a label) were being dropped from the resulting retrieve

Enable compression on all responses

Turn off some debug grid printing in logs

Fix issue when left axis starts with blank rows causing the connector to think that use both member names/aliases was turned on when it wasn’t, resulting in an error

Fix issues with blank result grids with suppress missing is turned on

Don’t mark label rows as invalid members

Fixed issue with how cell counts are calculated for purposes of breaking extremely large retrieves into multiple, smaller retrieves (cell count was not taking into account row member cells, only data cells)

Added optimization to prevent Member Selection dialog from taking a long time to open. #3875

Allow blank columns in the beginning of the "top" axis. #3982

Allow invalid member in first row of left axes. #3987

Enhanced support for keep/remove only operations so that they will work with multiple selections/ranges (however, these operations are currently based on row indices, not member names/groups)

Added clearCache=true parameter to cube/plan settings to force a complete cache clear (should only be used temporarily, e.g. set this parameter, then test the connection, then change it back)

Outline XML now indicates when members are shared. #3984

Outline XML now indicates attribute dimensions

Outline XML now supports the Default alias for members. #3971

Outline XML no longer shows additional member of same name as dimension under the dimension (e.g. Scenario had a child named Scenario, which has Actual, Budget, etc). #3972

Fix where selector list might be missing some hierarchies (was due to level computation issue). #3869

Enhanced retrieve/ad hoc operations to avoid the 500,000 cell limit in EPM Cloud by creating multiple requests and stitching them together

Requests to EPM Cloud will now use the current suppress missing option, which may reduce response time for retrieve operations

Blank rows are now allowed at the top of the "left" axes

Fixed issue where blank columns/rows would return "null" as cell value

Drill to all levels now honors ancestor on top option

Fixes to unknown member handling, support strict mode option

Fix zoom issue on horizontal/top axis

Drill to all levels returns members in correct order (parents on bottom as subtotals)

Fix issue with repeat member names option

Fix issue with Shares() DQE erroneously including parents

Introduce AllShares() DQE to expand all shares (can also nest Shares() calls)

Fix issue with using remove only on a top axis

Support Within Selected group ad hoc option

Support grids that result in no data (doesn’t cause error)

Support for Empty Grid Error option (to explicitly cause error when data grid is empty)

Reworked Zoom In to more closely mimic conventional zoom behavior

Fixes to "use both members and aliases" option

Adjustments to unknown member handling

Ships with updated Java Service Wrapper files, and includes support for macOS

Connector now returns connection properties to connection testing/troubleshooting in Dodeca 8.4+

Fixed issue with zoom operations returning members that aren’t in outline order

Added support for strict mode (Dodeca add-in option)

Added support for unknown members (list of unknown members is returned, retrieves skip invalid intersections unless strict mode is on)

Fixed a zoom issue when zooming on POV members

Added support for "use both members and aliases" option. Incoming grids may now provide a grid where each dimension appears twice (e.g. a row containing Jan, January) and the use both option will be honored in response grids

Adjusted indentation to only apply to rows but not members in the POV or headers

Fixed issue where selector lists may be missing first item in a hierarchy (such as Jan missing under Qtr1)

Performance optimization when querying a member that is a dimension

The "Ancestor on Top" option setting now defaults to false (the normal historical value), and is also now honored if set in retrieve options

Fixed issue where operations (such as zoom in) on a member that also has a share may result in incorrect results

Known issue: there is a potential issue with certain drill-to-bottom operations where an incomplete grid is coming back from the export data slice service

Member levels and generations are now included as part of the outline cache. This greatly speeds up indenting, zoom out, and other operations. If a member changes level/generation then the cache may need to be refreshed to reflect the change

Attribute dimensions may now be specified using the attributeDimensions key on a connection string (they may still be specified in the dimensions key). This allows the connector to exclude attribute dimensions from certain operations such as the default retrieve

Added basic support for remove only (rows)

Fixed issue when zooming when using a vertically stacked POV

Fixed issue where numeric values are being rounded to the nearest tenth of a dollar (e.g. the value 1000.65 may get rounded to 1000.70)

Fixed issue with dimensional query expressions (DQE) not working. DQEs now work correctly when used in a selector list with type EssbaseLegacyMemberQuery

Fixed issue when using Member Select dialog box

Fixed issue when a cached member selector referred to an alias name

Fixed issue when zooming to bottom or all levels

Added basic zoom out support

Retrieve operations now support SmartList values in the response grid

Fixed issue where querying for children/descendants of a member may return only the member itself if the member has a share in the outline

Added dimension query expression script feature to allow querying outline metadata for use in selectors and other contexts

Support for blank rows and columns in retrieve ranges

POV may start in top left corner of a retrieve range (POV must be horizontal)

Added support for launching a business rule using run time prompt variables

Fix various issues with grid metrics that were affecting drill-through and member selection

Support for indentation styles on grids

Support suppress missing and suppress zero in grids

Support navigate without data

Better support for using aliases in grids

Enhance zoom in, zoom to bottom, and multi-axis zoom

Fixed issue with member names starting with numbers in the POV

Fixed issue with caching member aliases and retrieving grid with aliases turned on

Fixed issue where cube/plan object was not being reused between requests

Member cache is now case-insensitive for server and member name Application and plan name are still case-sensitive: care should be taken to ensure consistent casing when defining connections in Dodeca and connections in the conncli.jar tool

Fixed issue where member was being looked up twice

Fix Windows service description

Added searchThreads parameter to database connection string to use the specified number of threads when performing member/alias search

Fix grid issue where additional requests to lookup member names was not going through cache.

Enhance conncli.jar to report more details on exceptions

The connector now includes an advanced caching architecture that is persistent between restarts and can be pre-populated using the conncli.jar tool.

You can now launch a business rule (calc) with a given name (uses the CalculateFromFile connector method)

Attribute dimensions are now properly supported in retrieve grids (make sure to add attribute dimension name to connection string, such as Basic?dimensions=Scenario,Account,…​,Market Size)

Added support for the connector to use a proxy when communicating with an Oracle EPM Cloud server

Upgraded version of PBJ library to 2.0.0