This release of the R package builds against TileDB 2.26.2, and has also been tested against earlier releases as well as the development version (#757)
Fix MacOS rpath (#760)
Fix "Can't read domain for dimensions of type UINT16" (#758)
Error messages displayed when a mismatched external pointer is detected now show both expected and encountered types (#740)
NDRectangle objects can now instantiate from more domain data types (#741, #742)
NDRectangle objects can now return their number of dimensions and dimension data types (#743)
FragmentInfo objects are dump via the << stringstream operator instead of a now-deprecated dump() method (#753)
The nighly valgrind matrix now includes release 2.26.0 (#744)
The continuous integration script has been updated reflecting external changes (#746)
as.data.frame, as.matrix and as.array to the tiledb_array() accessor, deprecated in release 0.20.0 in July 2023 in favor of the more general return_as="..." form, have been removed. (#751)as.data.frame it was temporarily re-admitted as an argument. It is expected to be removed following the upcoming 3.20 release of BioConducto.r (#752)Three deprecated calls to dump() methods for TileDB Embedded objects now use the preferred streaming alternatives (#727)
Two already deprecated functions that are removed in TileDB 2.26.0 are no longer used (#732)
The vendored nanoarrow has been updated to its release 0.5.0 (#733)
Fragments can now be removed by supplying a vector of fragment URIs (#734)
NDRectangle and CurrentDomain objects are supported (with 2.25.0 or newer) and can be used with ArraySchema and ArraySchemeEvolution domain of (#735, #737)
fromDataFrame, start and/or end timestamps can now be specified (#719)The nighly continuous integration matrix now included Core release 2.24.0 and 2.22.0 is dropped (#721)
The Conda build is now accomodating the change from #710 (#722)
Three internal and unexported helper functions now document more clearly how they can be called explicitly. (#709)
Reading and writing of text- and binary files supported by a VFS backend is now supported. (#710)
Building TileDB Embedded from source now uses tiledb install-tiledb as targets in a single CMake step. (#711, #713)
The time-travel tests now uses absolute (given) timestamps for writes as well as reads. (#716)
Group elements can now be deleted (#702)
Two error messages now show the human-readable type representation instead of the enum counter value (#705)
The test files receives a minor refactoring absorbing two files (#698)
The nightly valgrind run was updated to include release 2.23.0, release 2.21 has been removed (#703)
libtiledb_array_create_with_key, accessing a deprecated Core function, is now in src/deprecated.cpp and will be removed at later point (#699)libtiledb_query_add_range_with_type and libtiledb_query_add_range, deprecated in release 0.17.1 in January 2023, have been now removed (#700).The display of a filter_list is now labeled correctly as a filter list (@cgiachalis in #681 addressing #678)
The Arrow integration has been simplified using nanoarrow returning a single nanoarrow object; an unexported helper function nanoarrow2list() is provided to matching the previous interface (#682, #685)
An new accessor for recursive listings of (currently S3-only) URI is now available (with TileDB Core >= 2.22.0) (#691)
Initial support for TILEDB_GEOM_WKB and TILEB_GEOM_WKT has been added (with TileDB Core >= 2.21.0) (#692)
data.frame returns by tiledb_object_walk (#684 closing #683)The configure and Makevars.in received a minor update correcting small issues (#680)
The nightly valgrind run was updated to include release 2.22.0 (#687), release 2.19 and 2.20 have been removed (#695)
Functions tiledb_arrow_array_ptr(), tiledb_arrow_schmea_ptr(), tiledb_arrow_array_del() and tiledb_arrow_schema_del() are deprecated (in favor of using the corresponding nanoarrow functions) and will be removed in a future release (#685)
The function tiledb_query_submit_async() is marked as deprecated (as is the underlying C++ function) and slated for removal in a future release (#694)
The vendored nanoarrow sources have been update to release 0.4.0, and use of its facilities has been extended (#663)
Query conditions can be expressed against non-existing enumeration (i.e., factor) values when TileDB Core 2.21.0 or later is used (#674)
The tiledb_array_upgrade_version helper function to upgrade an schema version is now available (#675)
The tiledb_get_query_range_var() accessor now correctly calls the range getter for variable-sized dimensions (#662)
The nightly valgrind check now installs to require nanoarrow package (#664)
Variable cell numbers can now set consistently for all attribute types (#670)
Object walk traversal order detection has been corrected (#671)
The nightly valgrind run was updated to include release 2.21 (#669)
Unit tests have been added for the TileDB 'object' functions (#671, #672)
Obsolete checks for an ancient Windows version have been removed from the unit tests (#673)
Factor level additions now check for possible over in the index type (#645, #646)
Aggregate operations can now be performed on dense arrays via a query object with appropriate subarray settings (#650)
Factor level additions ensure the factor is releveled under the full set of factors (#644)
The example for fromDataFrame() has been updated, along with two other help files (#648)
Handling of temporary files in one test script has been standardized (#653)
A TileDB Array can now be opened in 'keep open' mode for subsequent use without re-opening (#630)
Arrays with factor (or ordered) variables now grow their factor levels in appending writes (#639)
Initialization of object walk order in recursive mode is now more explicit (#640)
Use of TileDB Embedded was upgraded to release 2.18.3 (#638), and 2.19.0 (#641)
Builds from TileDB Core non-release tarballs are now supported via new configure option (#627)
Tests are more careful about using suggested packages only when present (#632)
When building TileDB Core, shared linking is now requested explicitly (#634)
Nightly automated checks now include Core release-2.19 and add the 'curl' binary (#635)
Builds on maOS now set release 11 ('Big Sur') as the required minimum version (#636)
Use of TileDB Embedded was upgraded to release 2.18.0 (#620), 2.18.1 (#621), and 2.18.2 (#624)
Support for Aggregates has been added (#623)
Some tests were refactored slightly for greater robustness (#618)
Support for download and build with an external TileDB Core source tarball has been added (#622)
factor index columns other than the standard integer type (#614)Set conditions are supported in query condition expressions (#597)
Query conditions expression parsing via parse_query_conditions was extended simmilarly (#598)
Array fragment deletions uses a new static method (with TileDB 2.18.0 or later) (#599)
The included nanoarrow header and source file have been updated to release 0.3.0 (#600)
Query conditions expression parsing requirements are stated and tested more clearly (#601)
Use of TileDB Embedded was upgraded to release 2.17.2 (#602)
Enumeration (aka 'factor') support has been extended for 'empty' creation and subsequent extension with new levelss (#605)
Use of TileDB Embedded was upgraded to release 2.17.3 (#606)
Factor variables with (unlikely) int64 indices are supported (#607)
R objects can be (de-)serialized to and from VFS paths (#608)
Enumeration support has been extended to some cases only supported by Arrow (#609)
Use of TileDB Embedded was upgraded to release 2.17.4 (#611)
The DESCRIPTION file now correctly refers to macOS 10.14 (#596)
The (explicitly) 'batched reader now ensure a correct layout for sparse arrays (#610)
Array schema evolution has been extended to support enumerations (#590, #591)
Conversion to and from integer64 (and nanotime) now use package RcppInt64 (#592)
Use of TileDB Embedded was upgraded to release 2.17.1 (#593)
Use of TileDB Embedded was upgraded to release 2.17.0 (#583,#587)
Built-time configuration of TileDB Embedded can now be accessed as a JSON string (#584)
Enumeration types (i.e. what R calls factor variables) are now supported (#562)
Enumeration support has been extended to ordered types (#586)
The column buffer allocation is now robust to container overflow sanitizer checks (#574)
The array schema version is now accessible via a function (#575)
Use of TileDB Embedded was upgraded to release 2.16.1 (#576)
The tile extend getter function is now able to access a wider range of possible values (#577)
Several minor adjustments were made to not tickle any notives from R-devel CMD check especially on Windows (#571)
A memory allocation for Arrow objects was aligned with how that memory is later freed (#572)
The startup message is now reformated across two shorter lines (#545)
Support for returning 'Arrow Table' objects has been added (#548)
Use of TileDB Embedded was upgraded to release 2.15.3 (#551)
Date columns can now be exported to Arrow as well (#554)
Array writes which set timestamps now take advantage of the new temporal policy API (#558)
Displaying a schema is now more readable with additional linebreaks (#560)
Attempts to select on dimensions that do not exist now error out with a message (#561)
Integer64 values can now be written to array metadata (#564)
Date and POSIXct attributes are now supported in query conditions (#568)
Consolidation and vacuum calls now reflect the state of the global context object (#547)
Pointers to 'Arrow Table' objects representing the table columns are now in external pointers too (#550)
The documentation for 'Dimensions' was corrected in two spots in its wording / grammar (#552)
'sudo' mode is reenabled for package 'bspm' used in the continuous integration at GitHub Actions (#549)
Compilation on Linux systems as old as Ubuntu 18.04 without a filesystem header is now possible (#556)
as.data.frame, as.matrix and as.array to the tiledb_array() accessor are deprecated in favor of the more general return_as="..." form. (#567)timestamp argument to tiledb_array, deprecated in favor of timestamp_end (and timestamp_start) in July 2021, has been removed (#566).Query conditions can now be expressed for attributes of type UTF-8 (#529)
The startup message now displays the operating system and version (#532)
Use of TileDB Embedded was upgraded to release 2.15.1 and 2.15.2 (#534, #541)
Group objects can be opened while supplying a Config object when 2.15.1 or newer is used (#535, #536)
For character column buffer allocations, the R function now accepts a nullable option (#537)
For standard buffer allocations, the R function now accepts nullable and varnum options (#538)
Query conditions can now be expressed on boolean attributes (#540)
Testing for Groups reflect the stricter behavior in config setting requiring a close array (#530)
The use of binary packages in continuous integration has been made a little more robust (#531)
A small subset of tests are skipped if testing against the older release 2.14.* (#542)
r.legacy_validity_mode to true; the option also permits to write to an older installation. A conversion helper script is provided in scripts/legacy_validity_convert.r. (#517)Attributes can now be created, written and read from in (explicit) UTF8 types (and CHAR and ASCII already behaved correctly with respect to utf8 data) (#510)
Compilation under clang++ no longer complains about two unused member variables (#512)
Query conditions for character columns can now be expressed using the %in% operator and a vector of values (#513)
Use of TileDB Embedded was upgraded to releases 2.14.1 and 2.15.0 (#516, #521)
Safer checking of NAs in tiledb_config() to support R 4.2 conditional lengths (#519)
Query conditions can now be combined using & and | (in addition to && and ||) (#526)
The TileDB Embedded version is now used to determine whether a dampener is needed for the deprecation warning (#511)
One of the test data sets included with #517 has been regenerated under an older TileDB version in order to test on more systems (#523)
Documentation for Metadata accessors no longer states URIs strings are accepted (#527)
The nightly valgrind job matrix was updated to releases 2.13 and 2.14 as well as the branch (#504)
The nightly valgrind job show the most recent commit sha1 after updating from release branches (#507)
A query condition test for utf8 attributes has been added (#507, #508)
Support for Subarrays to set ranges has been extended (#496)
Deprecated Core API functions for Array access and range setting are longer used (#496)
TileDB Group objects now have a default show() method (#498, #499)
Domain and tile sizes for int64 dimension objects are now internally converted (#500)
Use of TileDB Embedded was upgraded to release 2.13.1 (#501)
Fragment info domain getters now work with ASCII domains (#495)
The scale filter option setting was corrected to use the proper types (#503)
The nightly valgrind job setup was updated to include two new dependencies (#493)
The Windows setup for continuous integration was updated (#494)
libtiledb_query_add_range{,_with_type} relying on depecreated Core
functionality are deprecated, and will be removed with the Core functions.
Subarray range setters are available. This is a mostly internal change.libtiledb_query_set_coordinates() and libtiledb_coords()
which have been deprecated since June 2000 have been removed. (#497)Support for testing group URIs on being relative has been added (#478)
Logging support at the R and C++ level has been added (#479, #487, #489)
Use of TileDB Embedded was upgraded to release 2.12.1, and 2.12.2 (#480, #481)
Sparse array queries via tiledb_array and '[]' access use an UNORDERED query layout (#488)
Use of TileDB Embedded was upgraded to release 2.13.0 (#490)
Support for selecting dimensions by discrete points has been added (#491)
Accomodate possible zero sized allocation estimates for attributes (#482)
Detect missing columns in a write-attempt with partial data (#483)
Update check-out action to version three suppressing a warning (#477)
Code Coverage reports are now generated and available at codecov.io (#484)
Small internal changes renaming two files and conditioning tests under two older releases (#485)
Several deprecated API entry points of TileDB Embedded are no longer used (#452, #453)
Support for DELETE queries has been added (requires TileDB Embedded 2.12.0 or later) (#455, #456)
Use of TileDB Embedded was upgraded to release 2.11.1, 2.11.2, and 2.11.3 (#460, #466, #474)
Support for XOR filters has been added (#472)
Support for deletion of fragments has been added (#473)
Use of TileDB Embedded was upgraded to release 2.12.0 (#475)
Treatment of character columns with missing values has been corrected (#454)
Accessing encrypted arrays has been reverted to the older API accessors (#458)
Int64 domain values in excess of int range are now expressed as integer64 objects (#465)
Sparse matrix conversion used mainly in tests have been updated for version 1.4-2 of the Matrix packages (#457)
Support builds on the riskv64 platform by adding a missing link instruction (#459)
The test setup was tweaked to not trigger a spurious valgrind report from libcrypto (#461)
The test setup was tweaked to make a group comparison more resilient to ordering (#462)
The test setup was refined for two filter tests (#467, #468)
A parameterized test for the SCALE_FLOAT filter has been added (#469)
The test setup ensures that the per-session directory remains accessible (#470)
Continuous integration testing for Linux and macOS has been moved to GitHub Actions (#471)
Support for query conditions has been extended to dense arrays (#447)
Support for filter lists has extended to both the data.frame helper and the dimension object constructor (#448)
Use of TileDB Embedded was upgraded to release 2.11.0 (#449)
Use of TileDB Embedded was upgraded to release 2.10.2 (#443) following an earlier update to 2.10.1 (#434)
List columns are now supported in reading and writing of data frames by extending cell variable numbers beyond one (#438, #440)
Query condition support has been extended to more data types (#441)
The 'SCALE_FLOAT' filter for compression of floating-point attributes is now supported (with TileDB 2.11 or later) (#445)
Unit tests were refined with some additional conditioning on envuironment variable CI being present (#436)
An unnessary final argument has been dropped from a 'remove member' method (#437)
Use of TileDB Embedded was upgraded to release 2.10.0 (#432) following earlier updates to 2.9.1 (#415), 2.9.2 (#419), 2.9.3 (#422), 2.9.4 (#427) and 2.9.5 (#430)
The BOOL data type is now supported (#416)
Query conditions support was extended with support for an OR operator (#417)
An incomplete query result is now signaled via a warning message (#420)
A helper function was added to check if an Array is open (#421)
Batched queries are now supported given the user the possibility to process larger-than-memory result sets in parts (#429)
Some internal object creation code was refactored (#431)
attr setter for Attributes was corrected to support NA settings (#425)Filter compression tests are skipped on systems lacking AVX2 support (#418)
The build system now checks for C++17 support (#424)
The valgrind test was upgraded to release 2.9.3 and the 2.10 release branch (#426)
Tests for overlapping ranges have been added (#428)
Support for groups has been added for TileDB 2.8 or later (#404)
The group member name retrieval can now also return the optional group member name (#399)
Allocation and creation of large string vector buffers was refactored (#400)
Support for dictionary encoding compression filters has been added for TileDB 2.9 or later (#404)
Support for Filestore functionality has been added for TileDB 2.9 or later (#410)
Support for BLOB datatypes has been added for TileDB 2.7 or later (#411)
Use of TileDB Embedded was upgraded to release 2.9.0 (#413) following earlier updates to 2.8.1 (#401), 2.8.2 (#403), 2.8.3 (#408)
check() function is now deprecated and schema_check() is provided (#409)Nightly valgrind checks were updated to use current versions (#397, #402)
Following release of R 4.2.0, only ucrt builds are supported on Windows (#405)
A schedule nightly continuous action now checks current and release-candidate branches of TileDB with the R package under valgrind (#387)
All external pointers are now tagged and validated at compile- and run-time (#389)
A now-redundant group-creation method has been removed (#391)
Unit tests for group member addition were added and updated (#393)
Group members can also be added or removed by name (#395)
Use of TileDB Embedded was upgraded to release 2.8.0 (#396) following an earlier upgrades to 2.7.0 (#372) and 2.7.1 (#384)
The detection of TileDB headers and library is now more robust for cases where pkg-config is present but does not know about TileDB (#385)
The package documentation website was updated (#386)
A fallback was added for external pointer creation to support compilation without group support in TileDB Embedded (#390)
An incorrectly specified function call was corrected (#392)
The templated initialization for external pointer is now inlined to satisfy all compilers (#394)
Use of TileDB Embedded was upgraded to release 2.6.4 (#384) following an earlier upgrade to 2.6.2 (#359)
Creations of arrays from data.frame objects now supports a mode= argument with values 'ingest', 'schema_only', and 'append' (#360)
Some unit test and continuous integration code was refactored (#364, #375)
Finalizer use is now simplified taking advantage of an Rcpp change (#366)
A new option strings\_as\_factors was added for data.frame retrieval (#367)
The arrow C-level interface now uses external pointer objects following Arrow 7.0 (#368)
Support for memory limits has been extended, and partial reads are using with iterations to complete (#371)
Fragment info reading now account for the __fragments object (#373)
A nightly test under valgrind has been added; results are reported to slack (#382, #383)
UTF-8 string in metadata are now supported (#377)
Attribute-less arrays can now be created, written, and read (#378), also via higher-level accessors (#379)
Array status is now checked before closing (#362)
Signed and unsigned int64 dimensions are now mapped correctly from 'square-bracket indexing', and the third dimension is recognised (#365)
Domain information could overflow int64_t if an unsigned value was used, this now flips to double (#370)
Unit tests for consolidation and vacuuming were update to account for __fragments too (#374)
A unit test was corrected to ensure logical expressions are of length one (#381)
A new vignette on data ingestion has been added (#357)
A new vignette on installation options has been added (#358)
The vignettes are now built using package simplermarkdown (#361)
Help pages were polished (#369)
tiledb_dense and tiledb_sparse functions which were deprecated in February 2021 have been removed after a twelve-month grace period.Use of TileDB Embedded was upgraded to release 2.6.1 (#354) following an earlier upgrade to 2.6.0 (#340)
A cell value getter for dimension was added (#341)
Getter and setter functions for validity filter lists have been added (#349)
Memory budget use has been refined via a configurable budget setting (#346, #350)
A context getter function was added for query objects (#351)
The schema display functionality was refactored and extended (#342, #343, #344, #345, #352, #355)
Use of TILEDB_CHAR is deprecated in favor of TILEDB_STRING_ASCII (#353)
A .nojekyll file was added to prevent unnecessary GitHub Pages builds (#339)
A getter for fill values is only called with TileDB 2.1.0 or later (#347)
GitHub Actions on Windows no longer install qpdf which was never used (#348)
The stopifnot() assertions now use consistent error messages across all functions (#331)
A helper function matching TileDB data types to R types is now exported (#336)
The boolean variable for 'nullable' is now set with a default value (#329)
A test for accessing shared memory segements is now correctly checking for TileDB 2.6.0 (#332)
CI tests were expanded to also test refactored TileDB Embedded readers (#310), and now deactivated as this is now part of release 2.5.0 (#321)
The minimal version of TileDB Embedded that can be used with the R package is now release 2.0.0 (#313)
The package now compiles using the C++17 standard just like TileDB Embedded (#314)
Shared-memory interprocess communication is used to accelerate operation for TileDB Cloud (#316)
The long-deprecated max_element_size function has been removed from TileDB Embedded, and the R interface was updated accordingly (#317, #319)
The extended toggle and field for tiledb_array() can now select dimension-less returns from sparse arrays (#318)
Use of TileDB Embedded was upgraded to release 2.5.0 (#321) following earlier upgrades to 2.4.2 and 2.4.3 (#308, #312)
A new quickstart example using the 'memory filesystem' was added, and one another example updated (#323)
Tests for time-traveling were refactored and now in a separate test file (#311)
Read-queries no longer call finalize() required only on write-queries (#309)
Some examples were updated with copy/paste corrections (#317)
Single-column data.frame returns work via drop=FALSE where needed (#320)
Accessors for context and query statistics were added (returning easily parseable JSON strings) (#293).
Initial support for schema evolution was added to add or drop attributes (#294).
Use of TileDB Embedded was upgraded to release 2.4.0 (#295)
Windows builds under GitHub Actions now also include the newer UCRT variant (#296).
The internal memory allocation has been switched to set_{data,offset,validity}_buffer functions (#297).
A convenience accessor for schema information retrieving 'dimension' or 'attribute' status has been added (#299).
The default array type fromDataFrame has been change to sparse to match some optimizations in TileDB Embedded, some unit tests have been updated accordingly (#300).
TileDB arrays can now be queried in expression using pipes (for row-wise filtering and colunb-wise selection) (#301).
When matrices as well as sparse matrices are written to arrays, optional row and column names are now supported as well (#303, #304).
The configure script was update to the standards of autoconf 2.69 as requested by CRAN (#305).
Use of set_{data,offset,validity}_buffer is made conditional on TileDB 2.4.0 or later to continue builds on older versions (#298).
Tests of piped expressions have been rewritten to be compatible with R versions earlier than 4.1.0 (#302).
A dangling documentation link in README.md was corrected, and another removed (#306).
When retrieving results via the [ operator, incomplete queries generate a warning (#283)
The interface to query element size of queries was extended (#282)
If query ends as 'incomplete', a warning is now issue (#283)
The status of the preceding query can now be accessed also when using a higher-level wrapper (#285)
Fragment Information can be accessed via high-level accessor functions (#286)
A preference for data type as which TileDB array data is returned can be set (#288)
Continuous Integration will use increased test coverage by installing more optional package (#289)
Use of TileDB Embedded was upgraded to release 2.3.4 (#290)
One cast statement was corrected so a warning is no longer triggered from clang (#281)
Some added unit tests were not conditional on TileDB Embedded 2.2.* or later (#284)
A time-comparison unit test did not properly respect timezones which was corrected (#287)
A query condition parser was added for use with standard (non-quoted) R expressions (#267)
Windows UCRT builds at CRAN are now supported (#268)
Use of TileDB Embedded was upgraded to release 2.3.2 (#270), and again to 2.3.3 (#280)
The vacuum and consolidation helper functions now use time stamp support (#271)
The time-travel array opening support was updated to start and end timestamps (#272)
Tests for both vacuuming and consolidation 'time traveling' were added (#273)
Nullable string string attribute support was improved (#274)
Ascii columns attribute support was added (#276)
The query parser heuristic was improved to cover ascii strings (#277)
Array opening uses improvemed array state consideration skippingg re-openings for better performance (#279)
Domain size information gathering has been corrected for uint32, uint64, and int64 attribute domains (#266)
Timesteps for time-traveling unit tests were adjusted to not trip up macOS continuous integration tests (#275)
String array buffer size calculation was corrected also allowing for all-string arrays (#278)
The build defaults to TileDB Embedded 2.3.* (unless another version is found during build, or explicitly selected) (#258, #264)
Query condition support is available for TileDB 2.3.0 or later, allowing (possibly multiple) numerical constraints on attributes (#261)
Multi-dimensional arrays can now be returned from (dense) arrays via a new option (#263)
The package is now natively supported on Arm64 ("M1") macOS system (#264)
Continuous integration at GitHub is now faster as suggested packages are no longer installed (#250)
Arrays can now be written incrementally via the higher-level replacement function arr[] <- obj (#251)
The default column layout for arrays written via fromDataFrame is now column-order (#254)
The call to vaccum not correctly calls the library function to vaccum instead of the consolidation function (#252)
When several columns are selected via selected_ranges, a potentially necessary reordering is done for a query (#253)
Dense subarrays can be written for tiledb_array (#256)
Matrix objects can now be returned under range selections (#247)
Matrix return get turned on and off with setter / getter functions (#248)
Unit tests of character columns in data frames accomodate R versions prior to R 4.0.0 in all cases (#243)
Dimension reduction for attribute-selected columns was incorrect in some cases (#245)
Attribute-selected columns were using incorrect dimension data types in some cases (#246)
A new vignette show use of TileDB array via RMariaDB and the MyTile extension to MariaDB (#221)
Matrices can now be returned directly from suitable two-dimensional TileDB arrays (#225)
More data types are supported in the non-empty domain accessor function (#229)
The DESCRIPTION, README.md and pkgdown site were updated (#230)
Creation of TileDB arrays from data.frame object has been made more robust (#238)
On startup, versions numbers of the R package and the TileDB Embedded library are displayed (#239)
The pkgdown website now shows the ChangeLog derived from this NEWS file (#240)
Two tests with datetime comparisons which fail only on one macOS system are now conditional (#216)
Result sets with all-character column now fall back to estimated result sizes (#217)
Setup of support for duplicate values in sparse arrays has been corrected (#223)
Error messages concerning an array types and selection mismatch are now clearer (#224)
Writes from data.frame objects to dense array revert back to column-major order (#226)
Tests of sparse writes to dense matrices now use UNORDERED layout (#228)
Data.frame returns of selected columns now coversion dimensions as well (#231)
Schema creation has been generalized and made more robust (#232)
Selection of dimension ranges now maps date and datetime values correctly (#233, #241)
Selection and setting of dimension ranges has been generalized and made more robust (#235, #236)
The older implementations tiledb_dense and tiledb_sparse are now marked as deprecated in favor of tiledb_array. No removal date is set or planned yet, but it is recommended to migrate to new code. (#180)
Updated the underlying TileDB library to use TileDB 2.1.2 on macOS and Linux (when no system library is found) (#181)
There is extended support for array creation directly from DataFrame objects. (#182)
Internal TileDB performance statistics can now be exported 'raw' in JSON format (for TileDB versions greater than 2.0.3). (#183, #186)
The vignette was updated with respect to the preferred used of tiledb_array. (#184)
The Hilbert cell layout added recently to TileDB Embbeded is supported. (#185)
Virtual File System functions now use the default VFS object which allows for a simpler interface (#187)
TileDB Array Dimension support has been extended to more data types (#188)
Datetime support for Dimensions and Attributes has been extended (#189)
The API coverage has been increased alongside an update of the documentation (#190)
The Array schema can now be accessed and printed directly from a URI (#191)
The accessor for the estimated array result size was updated with the updaded API (#192)
Initial Arrow support for importing and exporting vector has been added (#193, #199)
Support for the default TileDB Embedded library was set to 2.1.3 and 2.1.4 (#194, #195)
The package documentation was updated and extended (#196)
Support for 'time-traveling' access to arrays was extended (#197)
Support for the default TileDB Embedded library was set to 2.1.5 and 2.1.6 (#198, #200)
Support for the default TileDB Embedded library was set to 2.2.0 and 2.2.1 (#201, #202)
The vignettes were updated and extended (#203)
The fromDataFrame() function was extended (#204)
Some unit tests were conditioned on having TileDB Embedded 2.1.0 or later (#205)
Support for the default TileDB Embedded library was set to 2.2.2 and 2.2.3 (#206, #208)
The default TileDB Embedded library version is now set centrally (#207)
Support was added to write and read sparse matrices directly via higher-level functions (#209)
Arrow array and schema allocation and deallocation helper functions were added (#210)
Support for Nullable vectors has been added (#211)
Support for the default TileDB Embedded library has been set to 2.2.4 (#212)
Small tweaks to timezone and factor settings in unit tests, and vignette (#213, #214, #215)
The tiledb_stats_reset() function is now exported, and tiledb_stats_print() has been re-added as a wrapper to tiledb_stats_dump() (#174)
Configuration options for compute and input/output concurrency set only the new TileDB 2.1 configuration options; documentation on how to checking values has been expanded. (#175)
The download.file() use now (re-)sets the timeout to the standard value to accomodate uses where a lower value may be set such as some CRAN builders (#176)
Build scripts have been updated for use of TileDB 2.1.1 on Windows, macOS and Linux (when no system library is found) (#178)
R-based metadata accessors have been extended to also support tiledb_array arrays (#169)
configure now also checks the hardware platform before attempting a download of a prebuilt library (#170)
SystemRequirements: in the DESCRIPTION file has been expanded (#170)
A new function limitTileDBCores() controls resource use, it is being used in tests (#139)
The function tiledb_get_context() is now exported (#140)
A new S4 class tiledb_vfs provides access to the virtual file system functionality (#140)
Functionality of selected_ranges() was extended (#142)
More (signed and unsigned) integer types are supported as dimension types in sparse arrays (#143), as well as in dense arrays (#144) and as attributes (#144)
A new S4 class tiledb_query offers access to the query object functionality in the underlying library (#145, #161)
Examples are running with a lowered thread count setting per CRAN Policies (#152)
External pointer objects now use explicitly set finalizers (#149)
Users can explicitly select a TileDB Embedded shared library built to be used (#151)
Compile-time configuration was refactored and changes (#158)
Windows builds are now possible also using TileDB Embedded build 2.0.8 (#159, #164)
Continuous integration now uses Azure for macOS and Linux (#160) and GitHub Actions for Windows (#162,#165)
Downloading the prebuild library accomodates multitple targets per architecture (#150)
The number of TBB threads will only be set once (#158)
Range selection for tiledb_array objects can get/set matrices defining range (#132)
The show methods are now consistently exported and documented (#134)
TileDB is listed as copyright owner in DESCRIPTION as well (#134)
The selected_ranges method for tiledb_array types was improved, and more tests were added (#135)
C++ source code was rearranged slightly with respect to possible API deprecations in the libary (#136)
A very simple example for using TileDB Cloud from R was added (#136)
The helper scripts for the package build are now in the tools/ directory (#137)
The (optional) library download now relied on suggested R packages 'jsonlite' and 'curl' (#137)
All S4 classes are now consistently documented or aliased (#117)
If needed, the build system now builds TileDB and its required component (#118)
Data.frame support has been extended further and made more robust (#119, #123, #128)
The Description: in DESCRIPTION has been refreshed (#120)
Builds on Linux and macOS can use a pre-built TileDB library (#121, #122, #124, #127)
Copyright headers were added to source files (#125)
The pkg-config helper can be used when building from source (#126)
Added support for heterogenous domains
Added support for string dimensions
Added support for duplicate dimension values in sparse arrays
Added support for data.frame object import and conversion to dense and sparse arrays
Added enhanced support for data.frame returns from dense array
Added support for data.frame column selection (i.e. attributes) from dense array
Added support for new filter types for md5 and sha256 checksums
Added support for Date, Datetime (i.e. POSIXct) and nanosecond dense and sparse array attributes and domains
Documentation and examples were enhanced and extended
tiledb_stats_dump() is now simpler (but needs to be enabled first as
usual)
Support for nanosecond and integer64 columns was added, this is an optional feature for which the nanotime (and bit64) packages need to be installed
Added support for i) multi-range subarrays, ii) incomplete queries, iii) result size estimation and 'time travel' at to time-points has been added #105
Added additional support for metadata #106