rowproof 0.1.2__tar.gz → 0.1.4__tar.gz

{rowproof-0.1.2 → rowproof-0.1.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rowproof
-Version: 0.1.2
+Version: 0.1.4
 Summary: Prove a database migration or replication didn't lose, add or change a single row.
 License: Apache-2.0
 Requires-Python: >=3.11
@@ -73,6 +73,8 @@ public.orders  postgres:postgres/public.orders  ->  clickhouse:default/orders
   result   DIFFERENT   exit 1
 ```
 
+**Next:** the [user guide](docs/USAGE.md) covers connection strings for each database, every option with examples, checking many tables from a config file, using rowproof in CI, and troubleshooting.
+
 ## What it checks — and what it never does
 
 - Verifies every row matches: counts, missing rows, extra rows, and changed values, down to the specific column and the exact rule that governed the comparison (a timestamp precision mismatch and a NULL-vs-empty-string mismatch are not the same bug, and rowproof tells you which one you have).
@@ -114,7 +116,7 @@ Every cross-engine type difference is governed by one of these published rules,
 | BOOL-1 | boolean | `true` / `false` | Handles integer 0/1 vs boolean across engines |
 | TS-1 | timestamp with tz | ISO 8601 UTC, `YYYY-MM-DDTHH:MM:SS.ffffffZ` | Always 6 fractional digits |
 | TS-2 | timestamp precision differs | rounded to the pair's lower precision | e.g. Postgres µs vs ClickHouse `DateTime` (seconds) |
-| TS-3 | timestamp without tz | rendered as-is with `Z`, assumed UTC, warns once | `--assume-tz` overrides |
+| TS-3 | timestamp without tz | rendered as-is with `Z`, assumed UTC, warns once | Always UTC for now; `--assume-tz` other than UTC is rejected (not implemented yet) |
 | DATE-1 | date | `YYYY-MM-DD` | |
 | TIME-1 | time | `HH:MM:SS.ffffff` | |
 | UUID-1 | uuid | lower-case, hyphenated | Handles a uuid stored as VARCHAR (Snowflake) |

{rowproof-0.1.2 → rowproof-0.1.4}/README.md

@@ -51,6 +51,8 @@ public.orders  postgres:postgres/public.orders  ->  clickhouse:default/orders
   result   DIFFERENT   exit 1
 ```
 
+**Next:** the [user guide](docs/USAGE.md) covers connection strings for each database, every option with examples, checking many tables from a config file, using rowproof in CI, and troubleshooting.
+
 ## What it checks — and what it never does
 
 - Verifies every row matches: counts, missing rows, extra rows, and changed values, down to the specific column and the exact rule that governed the comparison (a timestamp precision mismatch and a NULL-vs-empty-string mismatch are not the same bug, and rowproof tells you which one you have).
@@ -92,7 +94,7 @@ Every cross-engine type difference is governed by one of these published rules,
 | BOOL-1 | boolean | `true` / `false` | Handles integer 0/1 vs boolean across engines |
 | TS-1 | timestamp with tz | ISO 8601 UTC, `YYYY-MM-DDTHH:MM:SS.ffffffZ` | Always 6 fractional digits |
 | TS-2 | timestamp precision differs | rounded to the pair's lower precision | e.g. Postgres µs vs ClickHouse `DateTime` (seconds) |
-| TS-3 | timestamp without tz | rendered as-is with `Z`, assumed UTC, warns once | `--assume-tz` overrides |
+| TS-3 | timestamp without tz | rendered as-is with `Z`, assumed UTC, warns once | Always UTC for now; `--assume-tz` other than UTC is rejected (not implemented yet) |
 | DATE-1 | date | `YYYY-MM-DD` | |
 | TIME-1 | time | `HH:MM:SS.ffffff` | |
 | UUID-1 | uuid | lower-case, hyphenated | Handles a uuid stored as VARCHAR (Snowflake) |

{rowproof-0.1.2 → rowproof-0.1.4}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "rowproof"
-version = "0.1.2"
+version = "0.1.4"
 description = "Prove a database migration or replication didn't lose, add or change a single row."
 readme = "README.md"
 requires-python = ">=3.11"

rowproof-0.1.4/rowproof/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.1.4"

{rowproof-0.1.2 → rowproof-0.1.4}/rowproof/cli/main.py

@@ -1,4 +1,4 @@
-"""rowproof CLI — argparse stand-in for `typer` (see docs/DEV_ENVIRONMENT.md).
+"""rowproof CLI — argparse stand-in for `typer` (see docs/internal/DEV_ENVIRONMENT.md).
 Flags, behavior and exit codes follow spec §7 exactly; only the argument
 *parsing library* differs from the spec's chosen stack.
 """
@@ -201,12 +201,29 @@ def resolve_algorithm(requested: str, args) -> str:
     return "hashdiff"
 
 
+def _check_assume_tz(value: str) -> str:
+    """`--assume-tz` is accepted by the parser but no connector applies it:
+    a naive (TS-3) timestamp is always rendered as if it were UTC. Passing
+    another zone used to be silently ignored while the warning still said
+    "assumed <zone>" -- rows that match in that zone were reported
+    DIFFERENT with a message claiming otherwise. Refuse loudly until the
+    conversion is actually implemented in each connector.
+    """
+    if value.strip().upper() != "UTC":
+        raise RowProofError(
+            f"--assume-tz {value} is not supported yet: naive timestamps are always "
+            "compared as UTC. Use timestamp-with-timezone columns, or convert the "
+            "column in a view, so both sides agree on the instant."
+        )
+    return value
+
+
 def _normalise_kwargs(args) -> dict:
     return dict(
         trim=args.trim,
         case_insensitive=args.case_insensitive,
         float_precision=args.float_precision,
-        assume_tz=args.assume_tz,
+        assume_tz=_check_assume_tz(args.assume_tz),
         column_map=_parse_column_map(args.column_map),
     )
 
@@ -237,7 +254,7 @@ def cmd_diff(args) -> int:
                 # while also asking to sample -- a clear error beats
                 # silently ignoring the flag.
                 raise RowProofError(
-                    "--sample/--sample-rows requires hashdiff (spec §4.3) -- "
+                    "--sample/--sample-rows needs the hash-based method -- "
                     "pass --algorithm hashdiff, or diff two different connections "
                     "so hashdiff is auto-selected"
                 )
@@ -391,7 +408,7 @@ def _run_one_job(job, connections: dict, verbose: bool) -> int:
             key_columns=job.key, columns=job.columns, exclude=job.exclude,
             where=job.where, where_source=job.where_source, where_target=job.where_target,
             trim=job.trim, case_insensitive=job.case_insensitive,
-            float_precision=job.float_precision, assume_tz=job.assume_tz, column_map=job.column_map,
+            float_precision=job.float_precision, assume_tz=_check_assume_tz(job.assume_tz), column_map=job.column_map,
         )
         if algorithm == "joindiff":
             result = run_joindiff(

{rowproof-0.1.2 → rowproof-0.1.4}/rowproof/cli/render.py

@@ -1,5 +1,5 @@
 """Terminal + JSON rendering — plain-text stand-in for `rich` (see
-docs/DEV_ENVIRONMENT.md). Behavior/content matches spec §8; only the
+docs/internal/DEV_ENVIRONMENT.md). Behavior/content matches spec §8; only the
 "tables and progress bars" styling is simplified.
 """
 

{rowproof-0.1.2 → rowproof-0.1.4}/rowproof/connectors/_chwire.py

@@ -3,7 +3,7 @@ driver.
 
 `ClickHouseConnector` in clickhouse.py calls only the functions below,
 never clickhouse_connect directly. This module used to be a hand-rolled
-stdlib `urllib` HTTP client (see git history / docs/DEV_ENVIRONMENT.md
+stdlib `urllib` HTTP client (see git history / docs/internal/DEV_ENVIRONMENT.md
 for why); clickhouse-connect replaces that with the real driver, which
 speaks the same HTTP interface but also handles response decoding,
 per-engine type mapping, and connection re-use itself instead of this

{rowproof-0.1.2 → rowproof-0.1.4}/rowproof/connectors/_pgwire.py

@@ -3,7 +3,7 @@
 `PostgresConnector` in postgres.py calls only the functions below, never
 psycopg directly, so the rest of the codebase is insulated from driver
 specifics. This module used to shell out to the `psql` CLI binary (see
-git history / docs/DEV_ENVIRONMENT.md for why); psycopg replaces that
+git history / docs/internal/DEV_ENVIRONMENT.md for why); psycopg replaces that
 with a real wire-protocol connection and hands back already-typed Python
 values (int, Decimal, datetime, bool, ...) instead of text that has to be
 guessed back into a type.

{rowproof-0.1.2 → rowproof-0.1.4}/rowproof/core/hashdiff.py

@@ -497,8 +497,9 @@ def _append_run_level_warnings(plan: _Plan, result: DiffResult) -> None:
     if naive_cols:
         result.warnings.append(
             TdWarning(
-                "TS-3: naive timestamp column(s) assumed "
-                f"{plan.options.assume_tz} ({', '.join(naive_cols)}) — use --assume-tz to override",
+                "TS-3: naive timestamp column(s) compared as UTC "
+                f"({', '.join(naive_cols)}) — if they are really in another zone, "
+                "use timestamp-with-timezone columns or convert them in a view",
                 rule=NormalisationRule.TS_3,
             )
         )

{rowproof-0.1.2 → rowproof-0.1.4}/rowproof.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rowproof
-Version: 0.1.2
+Version: 0.1.4
 Summary: Prove a database migration or replication didn't lose, add or change a single row.
 License: Apache-2.0
 Requires-Python: >=3.11
@@ -73,6 +73,8 @@ public.orders  postgres:postgres/public.orders  ->  clickhouse:default/orders
   result   DIFFERENT   exit 1
 ```
 
+**Next:** the [user guide](docs/USAGE.md) covers connection strings for each database, every option with examples, checking many tables from a config file, using rowproof in CI, and troubleshooting.
+
 ## What it checks — and what it never does
 
 - Verifies every row matches: counts, missing rows, extra rows, and changed values, down to the specific column and the exact rule that governed the comparison (a timestamp precision mismatch and a NULL-vs-empty-string mismatch are not the same bug, and rowproof tells you which one you have).
@@ -114,7 +116,7 @@ Every cross-engine type difference is governed by one of these published rules,
 | BOOL-1 | boolean | `true` / `false` | Handles integer 0/1 vs boolean across engines |
 | TS-1 | timestamp with tz | ISO 8601 UTC, `YYYY-MM-DDTHH:MM:SS.ffffffZ` | Always 6 fractional digits |
 | TS-2 | timestamp precision differs | rounded to the pair's lower precision | e.g. Postgres µs vs ClickHouse `DateTime` (seconds) |
-| TS-3 | timestamp without tz | rendered as-is with `Z`, assumed UTC, warns once | `--assume-tz` overrides |
+| TS-3 | timestamp without tz | rendered as-is with `Z`, assumed UTC, warns once | Always UTC for now; `--assume-tz` other than UTC is rejected (not implemented yet) |
 | DATE-1 | date | `YYYY-MM-DD` | |
 | TIME-1 | time | `HH:MM:SS.ffffff` | |
 | UUID-1 | uuid | lower-case, hyphenated | Handles a uuid stored as VARCHAR (Snowflake) |

rowproof-0.1.2/rowproof/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.1.2"