Every release, what changed, and why.
Next to each "Connect" button the tool now shows "Connected: SQL Server <version>, <n> databases" in green or "Connection failed" in red. Changing the server, user, password, or authentication resets the indicator to "not connected". The toolbar wraps on narrow windows so no button is cut off.
The report scope "all tables" now lists every table in the grid; tables that were not compared appear as skipped. "Ticked tables only" contains the tables that were compared and are still ticked at export time. Before, the GUI held results only for the tables chosen for the run, so the two scopes looked identical.
The scope is now determined at export time from the grid: ticked tables (default), all compared tables, visible tables (filter), or highlighted rows. The report header states the scope and the number of tables. The table filter without a dot now searches the table name only, where it previously also hit the schema name; with a dot (dbo.T_) it searches Schema.Table.
The report can cover only the selected (compared) tables or all tables including the unselected ones. In the GUI a selector next to "Export HTML" decides (default: selected tables only); on the command line, --report-all adds the unselected tables. The report header states the scope.
The table filter now hides non-matching rows instead of ticking checkboxes. Text without * matches as "contains", with * as a pattern, several patterns separated by commas; Enter in the filter box works like the button. "All" and "None" affect visible rows only, and ticks you have set are kept while filtering. The tool also creates a Start menu entry on first start if none exists.
Data comparison of identically named tables in two SQL Server databases, as GUI and command line, with an HTML report. Three stages: metadata, bucket hash (one scan per side), detail only for differing buckets. Rows are paired by primary key, unique index, or a key you set; tables without a key are compared as multisets. Columns are matched by name with type and scale normalization; rowversion, computed, and generated-always columns are excluded, further columns by pattern. NULL, empty string, case, and trailing spaces stay distinct. Cost per table is shown first, large tables are not preselected, and the report can mask values.