Add support for SQL Server working copy

sno/base_dataset.py

@@ -9,6 +9,10 @@ class BaseDataset(ImportSource):
     """
     Common interface for all datasets - mainly Dataset2, but
     there is also Dataset0 and Dataset1 used by `sno upgrade`.
+
+    A Dataset instance is immutable since it is a view of a particular git tree.
+    To get a new version of a dataset, commit the desired changes,
+    then instantiate a new Dataset instance that references the new git tree.
     """
 
     # Constants that subclasses should generally define.

sno/checkout.py

@@ -20,7 +20,7 @@
 
 def reset_wc_if_needed(repo, target_tree_or_commit, *, discard_changes=False):
     """Resets the working copy to the target if it does not already match, or if discard_changes is True."""
-    working_copy = WorkingCopy.get(repo, allow_uncreated=True)
+    working_copy = WorkingCopy.get(repo, allow_uncreated=True, allow_invalid_state=True)
     if working_copy is None:
         click.echo(
             "(Bare sno repository - to create a working copy, use `sno create-workingcopy`)"
@@ -378,7 +378,7 @@ def create_workingcopy(ctx, discard_changes, wc_path):
         wc_path = WorkingCopy.default_path(repo.workdir_path)
 
     if wc_path != old_wc_path:
-        WorkingCopy.check_valid_creation_path(repo.workdir_path, wc_path)
+        WorkingCopy.check_valid_creation_path(wc_path, repo.workdir_path)
 
     # Finished sanity checks - start work:
     if old_wc and wc_path != old_wc_path:

sno/clone.py

@@ -106,7 +106,7 @@ def clone(
 
     if repo_path.exists() and any(repo_path.iterdir()):
         raise InvalidOperation(f'"{repo_path}" isn\'t empty', param_hint="directory")
-    WorkingCopy.check_valid_creation_path(repo_path, wc_path)
+    WorkingCopy.check_valid_creation_path(wc_path, repo_path)
 
     if not repo_path.exists():
         repo_path.mkdir(parents=True)

sno/geometry.py

@@ -57,6 +57,16 @@ def with_crs_id(self, crs_id):
         crs_id_bytes = struct.pack("<i", crs_id)
         return Geometry.of(self[:4] + crs_id_bytes + self[8:])
 
+    @property
+    def crs_id(self):
+        """
+        Returns the CRS ID as it is embedded in the GPKG header - before the WKB.
+        Note that datasets V2 zeroes this field before committing,
+        so will return zero when called on Geometry where it has been zeroed.
+        """
+        wkb_offset, is_le, crs_id = parse_gpkg_geom(self)
+        return crs_id
+
     @classmethod
     def from_wkt(cls, wkt):
         return wkt_to_gpkg_geom(wkt)
@@ -290,17 +300,26 @@ def gpkg_geom_to_ogr(gpkg_geom, parse_crs=False):
     return geom
 
 
-def wkt_to_gpkg_geom(wkb, **kwargs):
-    ogr_geom = ogr.CreateGeometryFromWkt(wkb)
+def wkt_to_gpkg_geom(wkt, **kwargs):
+    """Given a well-known-text string, returns a GPKG Geometry object."""
+    if wkt is None:
+        return None
+
+    ogr_geom = ogr.CreateGeometryFromWkt(wkt)
     return ogr_to_gpkg_geom(ogr_geom, **kwargs)
 
 
 def wkb_to_gpkg_geom(wkb, **kwargs):
+    """Given a well-known-binary bytestring, returns a GPKG Geometry object."""
+    if wkb is None:
+        return None
+
     ogr_geom = ogr.CreateGeometryFromWkb(wkb)
     return ogr_to_gpkg_geom(ogr_geom, **kwargs)
 
 
 def hex_wkb_to_gpkg_geom(hex_wkb, **kwargs):
+    """Given a hex-encoded well-known-binary bytestring, returns a GPKG Geometry object."""
     if hex_wkb is None:
         return None
 

sno/init.py

@@ -369,7 +369,7 @@ def init(
 
     if repo_path.exists() and any(repo_path.iterdir()):
         raise InvalidOperation(f'"{repo_path}" isn\'t empty', param_hint="directory")
-    WorkingCopy.check_valid_creation_path(repo_path, wc_path)
+    WorkingCopy.check_valid_creation_path(wc_path, repo_path)
 
     if not repo_path.exists():
         repo_path.mkdir(parents=True)

sno/repo.py

@@ -177,7 +177,7 @@ def init_repository(
         repo_root_path = repo_root_path.resolve()
         cls._ensure_exists_and_empty(repo_root_path)
         if not bare:
-            WorkingCopy.check_valid_creation_path(repo_root_path, wc_path)
+            WorkingCopy.check_valid_creation_path(wc_path, repo_root_path)
 
         extra_args = []
         if initial_branch is not None:
@@ -224,7 +224,7 @@ def clone_repository(
         repo_root_path = repo_root_path.resolve()
         cls._ensure_exists_and_empty(repo_root_path)
         if not bare:
-            WorkingCopy.check_valid_creation_path(repo_root_path, wc_path)
+            WorkingCopy.check_valid_creation_path(wc_path, repo_root_path)
 
         if bare:
             sno_repo = cls._create_with_git_command(

sno/sqlalchemy.py

@@ -1,3 +1,8 @@
+import re
+import socket
+from urllib.parse import urlsplit, urlunsplit, urlencode, parse_qs
+
+
 import sqlalchemy
 from pysqlite3 import dbapi2 as sqlite
 import psycopg2
@@ -6,6 +11,7 @@
 
 from sno import spatialite_path
 from sno.geometry import Geometry
+from sno.exceptions import NotFound
 
 
 def gpkg_engine(path):
@@ -95,7 +101,62 @@ def _on_connect(psycopg2_conn, connection_record):
             geometry_type = new_type((r[0],), "GEOMETRY", _adapt_geometry_from_pg)
             register_type(geometry_type, psycopg2_conn)
 
+    pgurl = _append_query_to_url(pgurl, {"fallback_application_name": "sno"})
+
     engine = sqlalchemy.create_engine(pgurl, module=psycopg2)
     sqlalchemy.event.listen(engine, "connect", _on_connect)
 
     return engine
+
+
+CANONICAL_SQL_SERVER_SCHEME = "mssql"
+INTERNAL_SQL_SERVER_SCHEME = "mssql+pyodbc"
+
+
+def sqlserver_engine(msurl):
+    url = urlsplit(msurl)
+    if url.scheme != CANONICAL_SQL_SERVER_SCHEME:
+        raise ValueError("Expecting mssql://")
+
+    # SQL server driver is fussy - doesn't like localhost, prefers 127.0.0.1
+    url_netloc = re.sub(r"\blocalhost\b", _replace_with_localhost, url.netloc)
+
+    url_query = _append_to_query(
+        url.query, {"driver": get_sqlserver_driver(), "Application Name": "sno"}
+    )
+
+    msurl = urlunsplit(
+        [INTERNAL_SQL_SERVER_SCHEME, url_netloc, url.path, url_query, ""]
+    )
+
+    engine = sqlalchemy.create_engine(msurl)
+    return engine
+
+
+def get_sqlserver_driver():
+    import pyodbc
+
+    drivers = [
+        d for d in pyodbc.drivers() if re.search("SQL Server", d, flags=re.IGNORECASE)
+    ]
+    if not drivers:
+        drivers = pyodbc.drivers()
+    if not drivers:
+        raise NotFound("SQL Server driver was not found")
+    return sorted(drivers)[-1]  # Latest driver
+
+
+def _replace_with_localhost(*args, **kwargs):
+    return socket.gethostbyname("localhost")
+
+
+def _append_query_to_url(uri, new_query_dict):
+    url = urlsplit(uri)
+    url_query = _append_to_query(url.query, new_query_dict)
+    return urlunsplit([url.scheme, url.netloc, url.path, url_query, ""])
+
+
+def _append_to_query(existing_query, new_query_dict):
+    query_dict = parse_qs(existing_query)
+    # ignore new keys if they're already set in the querystring
+    return urlencode({**new_query_dict, **query_dict})