CHANGELOG.md

Kart changelog

Please note that compatibility for 0.x releases (software or repositories) isn't guaranteed. Kart is evolving quickly and things will change. However, we aim to provide the means to upgrade existing repositories between 0.x versions and to 1.0.

When adding new entries to the changelog, please include issue/PR numbers wherever possible.

Unreleased

• diff: Use orjson for faster JSON-Lines output. #1019
• Upgrade to PDAL 2.7 #1005
• Adds a --drop-empty-geometry-features option to kart export. #1007
• Adds diagnostic output to Kart when KART_DIAGNOSTICS=1 environment variable is set. #1013
• In kart merge --output-format=json, the message field is now null if the merge is a no-op. #1021

0.15.3

• Replaces minimal patches with delta-filters - a more general-purpose way of filtering parts (inserts, updates, deletes) of JSON diffs when not all parts are required. #998
• Remove automatic resolution of localhost before calling MSSQL driver. Fixes SQL Server instance names issue. #999
• Upgraded pyodbc for compatibility with the latest MSSQL driver.

0.15.2

• Adds a new command kart export which enables export of vector or tabular datasets to any OGR format. #992

0.15.1

• Prevented committing local changes to linked datasets. #953
• Fixes a bug where even datasets marked as --no-checkout are checked out when the working copy is first created. #954
• Fixes a bug where minor changes to auxiliary metadata (.aux.xml) files that Kart is supposed to ignore were preventing branch changes. #957
• Improved performance when the user filters a diff request to only show certain features. #962
• Clean up the empty directory if a clone fails outright. #963
• Fixes add-dataset to work for PostGIS working copies. #965
• Fixes kart import --link to not fetch any tiles when importing into a bare Kart repo. #970

0.15.0

Major changes

• First and foremost: linked datasets! Kart now supports linking a tile-based dataset to tiles that are hosted on S3, so that the tiles don't need to be hosted on a Kart LFS server. For more information, see Linked Datasets. #905

Other changes

• Native Apple Silicon builds are now available for macOS 12 and newer. #927
• Adds support for disabling the working-copy checkout of specific datasets using the commands kart import DATASET --no-checkout or kart checkout --not-dataset=DATASET, and re-enabling it using kart checkout --dataset=DATASET. #926
• Adds information on referencing and citing Kart to CITATION. #914
• Fixes a bug where Kart would misidentify a non-Kart repo as a Kart V1 repo in some circumstances. #918
• Improve schema extraction for point cloud datasets. #924
• Some tweaks to --dry-run output of Kart LFS commands. #932
• Now using Python 3.11 to build Kart, and vendored dependencies have been updated to newer versions. #933
• Adds support for importing point-cloud and raster tiles directly from an S3 URL. #940
• checkout: Smoother progressbar updates
• Add library support for WEBP & ZSTD compression with GeoTIFF raster datasets. #951

0.14.2

• Fixes a bug where the GPKG application ID is not being written in GPKG working copies - other software reading the GPKG may be confused by this. #902
• Fixes a bug where points could shift by small amounts while importing point-clouds if the tiles were converted to COPC. PDAL#4180

0.14.1

• Fixes a bug where Git subprocesses (such as git clone) don't prompt the user for credentials or to resolve SSH issues on Windows. #852
• Better protection against XSS in the HTML diff viewer. #884
• Speed-up: greatly reduces the time taken to update the index file after checking out LFS tiles to the working copy. #880
• Adds --with-dataset-types option to kart data ls command (and deprecates kart data version). #892
• Use journal_mode = WAL in annotations.db sqlite database. #898
• Slightly modifies the GPKG working copy format to avoid an incompatiblity with Global Mapper. #899

0.14.0

Major changes

• First and foremost: raster datasets! Kart now supports raster datasets, which track changes to a set of GeoTIFF files. To get started, see Raster Datasets #794

Other changes

• Fixes a bug where SIGINT (Ctrl+C) wasn't terminating Kart when using helper mode, affects macOS and Linux. #869
• Fixes a bug where reflink wasn't working on macOS, leading to larger size on disk for point cloud datasets. #876

0.13.0

Major fix: Corruption of Postgres NUMERIC values

⚠️ This release fixes a bug causing corruption of NUMERIC fields when read from Postgres databases. #863

If your repositories use the NUMERIC or NUMERIC(a,b) types and were imported/committed from a Postgres source or working copy, you may have lost data.

Specifically, any numeric values with a scale of zero (no digits after the decimal point) were affected. If the only numerics in your repositories were defined with a scale greater than zero, no data corruption should have occured.

For example, a value of 100 in a postgres database with the type NUMERIC(10, 0) or NUMERIC would have been imported as 1 - losing its trailing zeroes.

However, a value of 100.0 in a field declared as NUMERIC(10, 1) would not have been affected.

Other data types (integer, floating-point etc) are not affected, and repos sourced from other SQL databases (Geopackage, MSSQL, MySQL) were not affected.

Other changes

• Allow kart to import non-spatial tables from PostgreSQL databases which don't have postGIS extension installed. #728
• PostgreSQL working copy no longer requires PostGIS for aspatial datasets. #729
• Fixes a bug where the current spatial filter isn't stored in the filesystem working copy, affecting the spatial filtering of point cloud datasets. #833
• Fixes pthread_key leaks in a long running Kart process due to repeated loading and unloading of mod_spatialite. #838
• Fixes a bug where features are written to the working copy without their CRS identifier during kart resolve. #840
• Drop CI testing of amazonlinux:2, replace with amazonlinux:2023. #838
• Point clouds (or rasters): during import, require the user to choose if they want to make a "cloud-optimized" dataset or a "preserve-format" dataset. #842
• Point clouds (or rasters): require --no-convert-to-dataset-format to add new tiles to a dataset that can be committed as-is but only if the dataset's format is changed. Note --convert-to-dataset-format works as before. #842
• Allow kart to check out attached files into the file-based working copy.
• Fixes a bug where tab-completion wasn't working with helper mode, affects macOS and Linux. #851
• Fixes a bug where git-lfs (and potentially other subprocesses) wasn't working reliably with helper mode, affects macOS and Linux. #853
• import now informs the user to use add-dataset instead if they try to import a table that is already inside the working copy. #249
• Fixes a bug where datasets with no CRS couldn't be checked out into a Geopackage working copy. #855
• Upgrade libgit2 to v1.6.4, pygit2 to v1.12.1 (plus Kart-specific changes). #868

0.12.3

• Fix the issue where re-importing the same shapefile with --replace-existing would show spurious schema changes (and therefore row changes). #791
• Added support for committing new layers to a kart repository with add-dataset and the table name. #249
• Switched the crs84Extent field for point-cloud tiles to store a 4-point-polygons (instead of an axis-aligned envelope). #809
• Improved progress reporting when importing tiles, in line with working copy checkout.
• Added support for --diff-format option under kart show command to display different level of detail depending on the user's needs. #721
• Allow for automatically resolving primary key conflicts during a merge using kart resolve --renumber=(ours|theirs) #814
• Improved tile import performance for point-cloud (and eventually raster) by making it multithreaded. #818
• Fixed a bug where Kart would require user.name and user.email to be set, even when GIT_AUTHOR_EMAIL and similar variables were set in the environment. #812
• Update native dependencies. #820
• Drop CI testing of Debian Stretch/oldoldstable. #820

0.12.2

• Fix for issue with launching an editor while using helper mode, affects macOS and Linux. #788
• Added support for --list-untracked-tables option under kart status command to list the tables/datasets that are added to the repository but are not tracked by Kart.

0.12.1

• Modified point-cloud-import --replace-existing to reuse previously imported tiles, rather than reimport them, if tiles that have already been imported are imported again - potentially saving time and disk space. Note that point-cloud-import --update-existing already had this optimization. #783
• Reintroduced support for log --graph. #784
• Added checks to various commands to ensure user.name and user.email have been configured (ie to kart merge), to bring them in line with kart commit behaviour. #785

0.12.0

Major changes

• First and foremost: point cloud datasets! Kart now supports point cloud datasets which track changes to a set of LAS files. This required the addition of a file-system working copy (as opposed to the existing database working copy types), and integrating Git LFS to store the large files. To get started with point cloud datasets, see Point Cloud Datasets #565

Other changes

• Install tab completion using kart install tab-completion instead of kart config --install-tab-completion. #701
• "Primary key conflicts" - conflicts caused by reusing a primary key which is already assigned to a feature outside the spatial filter - are renamed to "spatial filter conflicts" for future proofing with other dataset types. Consequently, commit option --allow-pk-conflicts is renamed to --allow-spatial-filter-conflicts.
• Support shell detection for tab completion installation when running in helper mode. #704
• Bugfix: directing geojson diff output to the console didn't work in a multi-dataset repo, even if only one dataset had changed. #702
• Improve argument parsing for kart diff and kart show. #706
• Bugfix: don't allow kart merge to fast-forward if the user specifies a merge message. #705
• diff, show, create-patch and apply now handle attached files (as could already be created using kart commit-files) #583
• metadata.xml is now consistently treated as an attached file, rather than as a hidden "meta item". #583
• apply and meta set now support --amend to amend an existing commit
• Bugfix: don't allow kart diff to diff to/from HEAD if HEAD does not yet point to a commit.
• Improved progress reporting during working-copy checkout. #738
• Support for changing the primary key column of an existing dataset. #238
• Help for the user get the working copy back into a valid state if a crash or similar causes it to become out of sync with the Kart repo. #751
• Enable the background CLI helper on Linux & macOS in CI builds. The helper improves CLI performance significantly. #776
• Dropped support for a variety of kart log options that were previously being forwarded to Git. #508

0.11.5

Major changes

• Tab completion. Run kart config --install-tab-completion=auto to install. Supports Windows cmd, Windows PowerShell, macOS and Linux. #643
• Add kart helper which starts a long running process to reduce the overhead of Python startup.

Other changes

• Add man-style pagers instead of click's default help page on kart <command> --help. #643
• Changed format of feature IDs in GeoJSON output to be more informative and consistent. #135
• Fixed primary key issues for shapefile import - now generates an auto_pk by default, but uses an existing field if specified (doesn't use the FID). #646
• Add --with-dataset-types option to kart meta get which is informative now that there is more than one type of dataset. #649
• Support kart diff COMMIT1 COMMIT2 as an alternative to typing kart diff COMMIT1...COMMIT2 #666
• Add kart helper which starts a long running process to reduce the overhead of Python startup.
• the --num-processes option to init and import commands is now deprecated and does nothing. In most situations it offered no performance gain. #692
• Honour symlinks for shared libraries rather than including copies created by PyInstaller. #691
• Strip shared libraries on Linux to reduce package size. #691

0.11.3

• Added support for --output option to kart conflicts. #135
• Bugfix: Better error message on using kart conflicts -ogeojson for meta-item conflicts. #515
• Removed the older upgrade-to-tidy and upgrade-to-kart features which were only relevant to Sno (predecessor of Kart). #585
• Added support for --decorate and --no-decorate in kart log. #586
• Bugfix: Fixed a bug where creating a MSSQL working copy fails when there are large (~10KB) geometries. #617
• Bugfix: Fixed kart diff <commit-id> for a commit containing a dataset that has since been deleted using kart data rm. #611
• Add ext-run to provide an execution environment for prototyping ideas/extensions.
• Added more context about datasets to JSONL diffs. #624
• Fix kart branch -o json when branch HEAD is unborn or repo is empty - this would either fail outright or report that HEAD is on branch "null" which is not exactly true. #637

0.11.1

• Improve performance when creating a working copy in a spatially filtered repository (this was previously slower than in a non-filtered repository; now it is much faster) #561
• Added Sphinx documentation, built from the docs directory. #220

0.11.0

Major changes

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. Starting with Kart 0.11.0, only the features that are inside the specified spatial filter are downloaded during a clone. Spatial filter docs | #456

Other changes

• Expanded --output-format/-o to accept format specifiers; e.g. -o json:compact. kart log now accepts text formatstrings, e.g. -o text:%H #544
• Deprecated --json-style in favour of -o json:{style}
• diff: Added --add-feature-count-estimate=<accuracy> to json-lines diffs. This lazily inserts an estimate of the total number of features into the output stream. #543
• Bugfix: fixed errors with Postgres working copies when one or more datasets have no CRS defined. #529
• Bugfix: better error message when kart import fails due to multiple XML metadata files for a single dataset, which Kart does not support #547
• When there are two pieces of XML metadata for a single dataset, but one is simply a GDALMultiDomainMetadata wrapping the other, the wrapped version is ignored. #547
• Bugfix: fixed a bug preventing checkout -b NEW_BRANCH HEAD^ and similar commands from working
• Bugfix: fixed a bug where kart merge would fail in a shallow clone, in certain circumstances. #555

0.10.8

• More advanced filters for log, diff and commit: All of these now work:
  • Wildcard (*) filters for dataset names, e.g. kart diff -- *parcel*:meta:schema.json will show only schema changes for all datasets with parcel in their names. * by itself matches all datasets. #532
  • You can now output the history of individual features: kart log -- <dataset-name>:feature:<feature-primary-key>. #496
• kart clone now strips .git off the end of automatically-generated repo paths, if --bare is not specified. #540
• Simpler developer builds using CMake, see the contributing notes.
• Bugfix: fixed certificate verification errors when cloning HTTPS repositories on some Linux distributions. #541
• Bugfix: fixed the error when merging a commit where every feature in a dataset is deleted. #506
• Bugfix: Don't allow --replace-ids to be specified during an import where the primary key is changing. #521

0.10.7

Bugs fixed

• log: Fixed some regressions in 0.10.6 involving argument parsing:
  • range arguments (x..y or x...y) were handled incorrectly. #504
  • -n INTEGER caused an error #507
  • kart log @ was empty (@ is supposed to be handled as a synonym for HEAD) #510
• Auto-incrementing PK sequences now work in PostGIS working copies for table names containing the . character. #468

0.10.6

• Information about the current spatial filter is now shown in status. #456
• Added a specification for allowed characters & path components in dataset names - see Valid Dataset Names.
• New kart data rm command to simply delete datasets and commit the result #490
• Fix for #491 - make Kart more robust to manual edits to the GPKG working copy that don't leave the metadata exactly as Kart would leave it (such as by leaving unneeded table rows in gpkg_contents)
• Added minimal patches:
  • kart create-patch now supports --patch-type minimal, which creates a much-smaller patch; relying on the patch recipient having the HEAD commit in their repository #482
  • kart apply now applies both types of patch.
• kart log now accepts a -- marker to signal that all remaining arguments are dataset names. #498
• import from a Postgres or MSSQL source will no longer prepend the database schema name to the imported dataset path.
• Bugfix: Diffing between an old commit and the current working copy no longer fails when datasets have been deleted in the intervening commits.
• Bugfix: Existing auto-incrementing integer PK sequences are now overwritten properly in GPKG working copies. #468

0.10.5

• Fixed regressions in diff -o geojson since Kart 0.10.1 #487
• Removed kart show -o geojson #487
• Fix for #478 merge --dry-run raises error
• Fix for #483 diff error with Z/M geometries

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.

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

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

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

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.

0.8.0 (Last "Sno" release)

Breaking changes

• 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 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/DBSCHEMA #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
• 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

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

0.6.0

Major changes in this release

• Newly created Sno repositories no longer have git internals visible in the main folder - they are hidden away in a '.sno' folder. #147
• 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/DBSCHEMA #267
  • Read the documentation at docs/POSTGIS_WC.md
• 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

0.5.0

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

0.4.1

Packaging fix:

• packaging: Fix issue with broken git component paths in packages on macOS and Linux (#143)
• packaging: Exclude dev dependency in macOS package

Minor features / fixes:

• Added a sno meta get command for viewing dataset metadata (#136)
• merge, commit, init, import commands can now take commit messages as files with [email protected]. This replaces the sno commit -F option (#138)
• import: Added --table-info option to set dataset metadata, when it can't be autodetected from the source database (#139)
• pull, push, fetch, clone commands now show progress - disabled with --quiet (#144)
• import now works while on an empty branch (#149)

0.4.0

Major changes in this release

• Basic conflict resolution:
  • sno merge now puts the repo in a merging state when there are conflicts (#80)
  • Added sno conflicts to list conflicts (#84)
  • Added sno resolve (#101)
  • Added sno merge --continue (#94)
• Major improvements to sno import and sno init --import:
  • Can now import from postgres databases (#90)
  • Multiple tables can be imported at once (#118)
• Added sno show: shows a commit. With -o json generates a patch (#48)
• Added sno apply to apply the patches generated by sno show -o json (#61)

Minor features / fixes:

• add a changelog (here!)
• sno import enhancements (in addition to major changes above):
  • GPKG database paths no longer need GPKG: prefix
  • now takes table names as separate arguments
  • Added --primary-key=FIELD to override primary key field name
  • Added --message to customize the commit message
  • --list no longer requires a repository
• sno init --import enhancements:
  • imports are much faster (#55)
  • now imports all tables from database, doesn't allow table to be specified
• Many JSON output improvements:
  • JSON output is specified with -o json instead of --json (#98)
  • Added syntax highlighting to JSON output when viewed in a terminal (#54)
  • sno diff JSON output layout has changed - features are now flat objects instead of GeoJSON objects. This is much more compact (#71)
  • Added JSON output option for most commands
  • Added --json-style option to several commands to control JSON formatting (#70)
• sno diff:
  • a..b now refers to the same changes as sno log a..b (#116)
  • can now diff against tree objects, particularly the empty tree (#53)
  • can now view some subset of the changes by supplying filter args, ie [dataset[:pk]]
• sno commit:
  • can now commit some subset of the changes by supplying filter args, ie [dataset[:pk]] (#69)
• removed import-gpkg command; use import instead (#85)
• Error messages now go to stderr instead of stdout (#57)
• Error conditions now use exit codes to indicate different types of errors (#46)

0.3.1

• Sno is now available on Windows 🎉 (Windows 8.1+ / Server 2016+ (64-bit))
• Updates to continuous integration — installers/archives are now built and tested with every commit for every platform.
• For macOS users, a homebrew "tap" is now available: brew cask install koordinates/sno/sno
• Several bug fixes

0.3.0

Major changes in this release

• License: sno is now publicly available under the GPL open source license.
• Sno now has a website at sno.earth
• Standalone builds and packaging for Linux and macOS. Windows won't be far behind
• Refactoring to support alternative database working copies
• Dependency upgrades (GDAL; Git; Pygit2; Proj; Libgit2; Sqlite; and others)
• Several bug fixes

Upgrading

If you were running a preview release, remove it before installing the new release:

$ brew uninstall sno
$ brew untap koordinates/sno

Repository Hosting

We have an initial preview available of our Sno repository hosting. This allows you & your team to push and pull Sno repositories. Please contact [email protected] with your Github username and we can get you set up. There is no cost for this service.

Compatibility

Repositories created with Sno v0.2 are compatible with v0.3. For assistance upgrading any v0.1 repositories, please read our upgrade guide.

0.2.0

Major changes in this release

• First and foremost, the name — we're now called Sno!
• A new repository structure/layout, which has better performance and a smaller on-disk size
• Data imports are now orders of magnitude faster
• Support for multiple datasets in a single Sno repository
• Support for non-spatial datasets
• Increased test coverage including end-to-end tests
• Improved macOS Homebrew packaging with CI testing.
• Prototype support for spatial-indexing and a sno query command for spatial lookups.
• Diffs across branches/commits, and a GeoJSON diff format.
• Numerous bug fixes