v0.10.4

0.10.4

Major changes

• Added basic support for spatial filters - the spatial filter can be updated during an init, clone or checkout by supplying the option --spatial-filter=CRS;GEOMETRY where CRS is a string such as EPSG:4326 and GEOMETRY is a polygon or multigon specified using WKT or hex-encoded WKB. When a spatial filter is set, the working copy will only contain features that intersect the spatial filter, and changes that happened outside the working copy are not shown to the user unless specifically required. #456

Other changes

• Auto-incrementing integer PKs: When the working copy is written, Kart now sets up a sequence which supplies the next unassigned PK value and sets it as the default value for the PK column. This helps the user find the next unassigned PK, which can be non-obvious in particular when a spatial filter has been applied and not all features are present in the working copy. #468
• Bugfix: Set GDAL and PROJ environment variables on startup, which fixes an issue where Kart may or may not work properly depending on whether GDAL and PROJ are appropriately configured in the user's environment
• Bugfix: kart restore now simply discards all working copy changes, as it is intended to - previously it would complain if there were "structural" schema differences between the working copy and HEAD.
• Bugfix: MySQL working copy now works without a timezone database - previously it required that at least UTC was defined in such a database.
• Feature-count estimates are now more accurate and generally also faster #467
• kart log now supports output in JSON-lines format, so that large logs can be streamed before being entirely generated.

v0.10.2

0.10.2

• Added support for the geometry POINT EMPTY in SQL Server working copy.
• Bugfix: fixed the error when writing diff output to a file. #453
• Bugfix: when checking out a dataset that has an integer primary key as a GPKG working copy, Kart should continue to use the actual primary key instead of overriding it, even if the primary key column isn't the first column. #455

Note on 0.10.0

Kart v0.10.0 introduces a new repository structure, which is the default, dubbed 'Datasets V3'. Datasets V2 continues to be supported, but all newly created repos are V3 going forward. See CHANGELOG.md or the previous release notes for more details about 0.10.0

v0.10.1

0.10.1

Fix for kart upgrade

Fixed kart upgrade so that it preserves more complicated (or yet-to-be-released) features of V2 repos as they are upgraded to V3. #448

Specifically:

• generated-pks.json metadata, extra metadata found in datasets that have an automatically generated primary key and which are maintained by repeatedly importing from a primary-key-less datasource
• attachments (which are not yet fully supported by Kart) - arbitrary files kept alongside datasets, such as license or readme files.

Other changes

• kart show now supports all the same options as kart diff. Both kart diff and kart show now both support output in JSON-lines format, so that large diffs can be processed as the diff is generated.
• Bugfix: diffs containing a mixture of primary key types can now be shown (necessary in the case where the primary key type has changed).
• Some performance improvements - less startup overhead.

0.10.0

Note: Kart v0.10.0 introduces a new repository structure, which is the default, dubbed 'Datasets V3'. Datasets V2 continues to be supported, but all newly created repos are V3 going forward. See CHANGELOG.md or the previous release notes for more details about 0.10.0

v0.10.0

Kart v0.10.0 introduces a new repository structure, which is the default, dubbed 'Datasets V3'. Datasets V2 continues to be supported, but all newly created repos are V3 going forward.

Datasets V3

• Entire repositories can be upgraded from V2 to V3 using kart upgrade EXISTING_REPO NEW_REPO.
• Anything which works in a V2 repo should work in a V3 repo and vice versa.
• V3 repos are more performant for large datasets - compared to V2 repos where size-on-disk climbs quickly once dataset size exceeds 16 million features.

Other major changes in this release

• The working copy can now be a MySQL database (previously only GPKG, PostGIS and SQL Server working copies were supported). The commands init, clone and create-workingcopy now all accept working copy paths in the form mysql://HOST/DBNAME #399
  • Read the documentation at docs/MYSQL_WC.md
• Import of tables using kart import is now supported from any type of database that Kart also supports writing to as a working copy - namely, GPKG, PostGIS, SQL Server and MySQL.
• Support for rapidly calculating or estimating feature-counts - see below.

Other minor changes

• Change to kart data ls JSON output, now includes whether repo is Kart or Sno branded.
• Importing from a datasource now samples the first geometry to check the number of dimensions - in case the datasource actually has 3 or 4 dimensions but this fact is not stored in the column metadata (which is not necessarily required by all source types). #337
• Bugfix: Creating a working copy while switching branch now creates a working copy with the post-switch branch checked out, not the pre-switch branch.
• Bugfix: GPKG spatial indexes are now created and deleted properly regardless of the case (upper-case or lower-case) of the table name and geometry column.
• A few bugfixes involving accurately roundtripping boolean and blob types through different working copy types.
• Bugfix: 3D and 4D geometries are now properly roundtripped through SQL Server working copy.
• Fix help text for discarding changes to refer to kart restore instead of kart reset, as kart restore is now the simplest way to discard changes. #426
• import: PostGIS internal views/tables are no longer listed by --list or imported by --all-tables, and can't be imported by name either. #439
• upgrade no longer adds a main or master branch to upgraded repos.

Calculating feature counts for diffs

Kart now includes ways to calculate or estimate feature counts for diffs. This encompasses the following changes:

• diff now accepts --only-feature-count=<ACCURACY>, which produces a feature count for the diff.
• log now accepts --with-feature-count=<ACCURACY> which adds a feature count to each commit when used with -o json.
• All calculated feature counts are stored in a SQLite database in the repo's .kart directory.
• Feature counts for commit diffs can be populated in bulk with the new build-annotations command

v0.9.0

0.9.0 (First "Kart" release)

Major changes in this release

• First and foremost, the name — we're now called Kart!

Other changes

• Various local config and structure which was named after sno is now named after kart - for instance, a Kart repo's objects are now hidden inside a .kart folder. Sno repos with the older names will continue to be supported going forward. To modify a repo in place to use the kart based names instead of the sno ones, use kart upgrade-to-kart PATH.
• import & init are often much faster now because they do imports in parallel subprocesses. Use --num-processes to control this behaviour. #408
• status -o json now shows which branch you are on, even if that branch doesn't yet have any commits yet.

v0.8.0

Breaking changes in this release

• Internally, Sno now stores XML metadata in an XML file, instead of nested inside a JSON file. This is part of a longer term plan to make it easier to attach metadata or other files to a repository in a straight-forward way, without having to understand JSON internals. Unfortunately, diffing commits where the XML metadata has been written by Sno 0.8.0 won't work in Sno 0.7.1 or earlier - it will fail with binascii.Error
• Backwards compatibility with Datasets V1 ends at Sno 0.8.0 - all Sno commands except sno upgrade will no longer work in a V1 repository. Since Datasets V2 has been the default since Sno 0.5.0, most users will be unaffected. Remaining V1 repositories can be upgraded to V2 using sno upgrade EXISTING_REPO NEW_REPO, and the ability to upgrade from V1 to V2 continues to be supported indefinitely. #342
• sno init now sets the head branch to main by default, instead of master. To override this, add --initial-branch=master
• reset now behaves more like git reset - specifically, sno reset COMMIT stays on the same branch but sets the branch tip to be COMMIT. #60
• import now accepts a --replace-ids argument for much faster importing of small changesets from large sources. #378

Other major changes

• The working copy can now be a SQL Server database (previously only GPKG and PostGIS working copies were supported). The commands init, clone and create-workingcopy now all accept working copy paths in the form mssql://[HOST]/DBNAME/SCHEMA #362
  • Currently requires that the ODBC driver for SQL Server is installed.
  • Read the documentation at docs/SQL_SERVER_WC.md
• Support for detecting features which have changed slightly during a re-import from a data source without a primary key, and reimporting them with the same primary key as last time so they show as edits as opposed to inserts. #212

Minor changes

• Optimised GPKG working copies for better performance for large datasets.
• Bugfix - fixed issues roundtripping certain type metadata in the PostGIS working copy: specifically geometry types with 3 or more dimensions (Z/M values) and numeric types with scale.
• Bugfix - if a database schema already exists, Sno shouldn't try to create it, and it shouldn't matter if Sno lacks permission to do so #391
• Internal dependency change - Sno no longer depends on apsw, instead it depends on SQLAlchemy.
• init now accepts a --initial-branch option
• clone now accepts a --filter option (advanced users only)
• show -o json now includes the commit hash in the output
• import from Postgres now uses a server-side cursor, which means sno uses less memory
• Improved log formatting at higher verbosity levels
• sno -vvv will log SQL queries to the console for debugging

v0.7.1

0.7.1

JSON syntax-highlighting fix

• Any command which outputs JSON would fail in 0.7.0 when run in a terminal unless a JSON style other than --pretty was explicitly specified, due to a change in the pygments library which Sno's JSON syntax-highlighting code failed to accomodate. This is fixed in the 0.7.1 release. #335

0.7.0

Major changes in this release

• Support for importing data without a primary key. Since the Sno model requires that every feature has a primary key, primary keys are assigned during import. #212
• Support for checking out a dataset with a string primary key (or other non-integer primary key) as a GPKG working copy. #307

Minor features / fixes:

• Improved error recovery: Sno commands now write to the working copy within a single transaction, which is rolled back if the command fails. #281
• Dependency upgrades (GDAL; Git; Pygit2; Libgit2; Spatialite; GEOS) #327
• Bugfixes:
  • sno meta set didn't allow updates to schema.json
  • Fixed a potential KeyError in Schema._try_align
  • Fixed a potential unexpected NoneType in WorkingCopy.is_dirty
  • Imports now preserve fixed-precision numeric types in most situations.
  • Imports now preserve length of text/string fields.
  • Imported fields of type numeric now stored internally as strings, as required by datasets V2 spec. #325

v0.7.0

Major changes in this release

• Support for importing data without a primary key. Since the Sno model requires that every feature has a primary key, primary keys are assigned during import. #212
• Support for checking out a dataset with a string primary key (or other non-integer primary key) as a GPKG working copy. #307

Minor features / fixes:

• Improved error recovery: Sno commands now write to the working copy within a single transaction, which is rolled back if the command fails. #281
• Dependency upgrades (GDAL; Git; Pygit2; Libgit2; Spatialite; GEOS) #327
• Bugfixes:
  • sno meta set didn't allow updates to schema.json
  • Fixed a potential KeyError in Schema._try_align
  • Fixed a potential unexpected NoneType in WorkingCopy.is_dirty
  • Imports now preserve fixed-precision numeric types in most situations.
  • Imports now preserve length of text/string fields.
  • Imported fields of type numeric now stored internally as strings, as required by datasets V2 spec. #325

v0.6.0

Major changes in this release

• The working copy can now be a PostgreSQL / PostGIS database (previously only GPKG working copies were supported). The commands init, clone and create-workingcopy now all accept working copy paths in the form postgresql://[HOST]/DBNAME/SCHEMA #267
  • Read the documentation at docs/POSTGIS_WC.md
• Newly created Sno repositories no longer have git internals visible in the main folder - they are hidden away in a '.sno' folder. #147
• Patches that create or delete datasets are now supported in Datasets V2 #239

Minor features / fixes:

• apply and import no longer create empty commits unless you specify --allow-empty #243, #245
• apply can now apply patches to branches other than HEAD #294
• apply, commit and merge commands now optimise repositories after committing, to avoid poor repo performance. #250
• commit now checks that the diff to be committed matches the schema, and rejects diffs that do not - this is possible in working copy formats that have relatively lax type enforcement, ie GPKG #300
• Added GPKG support for Sno types that GPKG doesn't support - they are approximated as strings. #304
• schema.json no longer stores attributes that are null - a missing attribute has the same meaning as that attribute being present and null. #304
• data ls now accepts an optional ref argument
• meta get now accepts a --ref=REF option
• clone now accepts a --branch option to clone a specific branch.
• switch BRANCH now switches to a newly created local branch that tracks BRANCH, if BRANCH is a remote branch and not a local branch #259
• gc command added (delegates to git gc)
• Bugfix - don't drop the user-supplied authority from the supplied CRS and generate a new unrelated one. #278
• Bugfix - generated CRS numbers are now within the user range: 200000 to 209199 #296

v0.5.0

sno v0.5 introduces a new repo layout, which is the default, dubbed 'Datasets V2'

Existing commands are backward compatible with V1 datasets, however some new functionality is only supported in repositories upgraded to the new layout.

Datasets V2

• Entire repositories can be upgraded from V1 to V2 using sno upgrade EXISTING_REPO NEW_REPO.
• V2 should support everything V1 supports
• All new repositories use the new layout by default. To opt out, use the --repo-version=1 flag for sno init
• A future release will drop support for v1 repositories

New features for V2 repositories only

• Most schema changes now work
  • this includes column adds, drops, renames and reordering.
  • Notably, changing the primary key field of a dataset are not yet supported.
• Meta changes are now supported (title, description and XML metadata for each dataset)
• import now has a --replace-existing flag to replace existing dataset(s).

Missing functionality in Datasets V2

• String primary keys and tables without primary keys are not yet supported. #212
• Changing the primary key column is not yet supported. #238
• Patches which create or delete datasets are not supported. #239
• Schema changes might not be correctly interpreted if too many changes are made at once (eg adding a new column with the same name as a deleted column - sno may incorrectly assume it is the same column).
  • It is safest to commit schema changes to any existing columns, then commit schema changes adding any new columns, then commit any feature changes.

Breaking changes in this release

• New structure to sno diff output:
  • Text output: Features are now labelled as <dataset>:feature:<primary_key>, consistent with meta items that are labelled as <dataset>:meta:<meta_item_name>
  • JSON output also uses "feature" and "meta" as keys for the different types of changes, instead of "featureChanges" and "metaChanges".
• sno show -o json header key changed to sno.show/v1, which is not an applyable patch. Use sno create-patch to create a patch.
• sno upgrade now only takes two arguments: sno upgrade EXISTING_REPO NEW_REPO. No other arguments are required or accepted, exactly how to upgrade the repository is detected automatically.

Other changes in this release

• Added sno create-patch <refish> - creates a JSON patch file, which can be applied using sno apply #210
• Added sno data ls - shows a list of datasets in the sno repository #203
• sno help [command] is a synonym for sno [subcommand] --help #221
• sno clone now support shallow clones (--depth N) to avoid cloning a repo's entire history #174
• sno log now supports JSON output with --output-format json #170
• sno meta get now prints text items as text (not encoded as JSON) #211
• sno meta get without arguments now outputs multiple datasets #217
• sno diff and sno show now accept a --crs parameter to reproject output #213
• Streaming diffs: less time until first change is shown when diffing large changes. #156
• Working copies are now created automatically. #192
• Commands which are misspelled now suggest the correct spelling #199
• Bugfix: operations that should immediately fail due to dirty working copy no longer partially succeed. #181
• Bugfix: some column datatype conversion issues during import and checkout.
• Linux: Add openssh client dependency into rpm & deb packages. #121
• Windows: Fix missing PROJ data files in packages. #235

External contributors

• OrangeOranges - better schema diffs, etc