Changelog

All notable changes to mgtx-xlwings-utils are documented here. The format is based on Keep a Changelog, and this project aims to follow Semantic Versioning.

[0.2.1] — 2026-07-10

Added

• offer_file_actions — the post-save “Open file / Open folder” dialog is now a public function (the internal _saved_dialog name still works as an alias), with a new Copy file path button and an actions= argument selecting which buttons to show (any subset of "open_file", "open_folder", "copy_path"; default all three). Copy file path copies the full path to the clipboard and leaves the dialog open so you can still open the file or its folder; those buttons (and OK) close it. Call it directly after saving a file any other way.
• save_dataframe(..., actions=...) — forwards the same button selection to the post-save dialog, so a batch step can, e.g., only offer "copy_path".

Fixed

• Copying the path now persists after the dialog closes. The button copies via the OS clipboard (PowerShell Set-Clipboard, UTF-8) instead of Tkinter’s delayed-render clipboard, which silently dropped the value when the window was destroyed.

[0.2.0] — 2026-07-06

A formatting + dropdowns release, plus plate and paste plumbing changes. Several breaking renames/removals are grouped under Changed.

Added

• apply_table_features — one call to apply column colour fills, prefill rows, and in-cell dropdowns to any managed table (manual or Benchling). Call it after the table’s init function. On an otherwise-empty table it auto-seeds a data row so a colour/dropdown has a cell to attach to.
• apply_benchling_table_features — the Benchling twin: hand back the results from initialise_table_benchling and every schema enum column becomes an in-cell dropdown automatically, seeded from the option list already resolved onto the result (no second Benchling call).
• paste_plate_data — refresh an already-initialised plate’s values without redrawing the grid, borders, merges, or named ranges (split out from create_plate_layout). Formatting and the user’s active tab survive untouched.
• paste_df_to_named_range(..., borders="mesh") — draw the cell-grid border mesh over the pasted extent.
• Format-preserving paste. paste_df_to_named_range now reads a formatted table’s template row back and propagates its column colours and dropdown validations across the pasted rows, rejecting off-list values in a dropdown column.

Changed

• Breaking — apply_benchling_dropdowns renamed to apply_benchling_table_features (it now does colours as well as dropdowns).
• Breaking — default table formatting is now the two-tone purple theme, applied through the shared apply_table_features path.
• Breaking — paste_df_to_named_range rejects a named (or MultiIndex) DataFrame index instead of silently dropping it; reset or rename it into a column first (#19).
• Breaking — plate per-column named ranges collapsed from O(rows × vars) to O(vars) and the dead per-column plate named ranges were dropped (#21). Plate reinit now purges orphaned variable names.
• Breaking — the dead paste(dropdowns=) parameter was removed; declare dropdowns via apply_table_features / apply_benchling_table_features.

Fixed

• Dropdown matching no longer trips over the formula-safety escape apostrophe, and the view stays on the table while dropdowns are built.
• Single-letter Benchling schema anchors no longer alias a column (#11).
• Source docstrings/comments are ASCII-only, with a regression guard (#12).
• Integration-test artefacts are written off the OneDrive sync path so the live-Excel suite no longer hangs mid-run (#13).
• Whole-column/row named ranges no longer hang a paste. The danger-zone scan tests each name’s extent rectangle instead of materialising every covered cell, so a sheet carrying a =$A:$A-style name pastes instantly.
• Paste-mode re-runs stack correctly. initialise_table_benchling(mode="paste") advances past a reinitialised table’s real extent, so a newly-added schema stacks below it instead of colliding at the anchor (TableSpaceOccupiedError).
• A first paste larger than the 1000-row scan window no longer inserts ~n - 1000 spurious blank rows below the table.
• Per-table dropdown option lists no longer clobber each other — two tables that share a column header (e.g. both “Status”) get separate namespaced option columns, and dropdowns on a sheet whose name contains an apostrophe (e.g. O'Brien) round-trip correctly.
• check_case_sensitive_entries compares by case-insensitive equality, not a regex substring match, so "Plasmid-derived" no longer false-matches "Plasmid" and a value like "C++" no longer raises re.error.
• get_named_range_results(strip=True) is shape-preserving — 1-D length and 2-D shape are kept (blank/None cells survive) so positional reads stay aligned.
• import xlwings_package no longer hijacks the host matplotlib backend; matplotlib.use("Agg") is deferred into paste_plot_to_cell, which also sorts a numeric x-axis numerically and rejects non-positive values on a log axis.
• A schema id listed twice for one sheet raises ValueError instead of silently writing the table twice and clobbering the first’s named ranges.
• Clearer errors on edge inputs: a zero-column DataFrame raises EmptyDataError; delete_sheet on a missing sheet raises WorksheetNotFoundError; scan_named_range_table anchored on row 1 raises NamedRangeNotFoundError.

Performance

• Plate draw and paste use bulk range writes instead of per-cell writes (#9).

Internal

• Deduplicated shared helpers (name-token extraction, column limits, title styling, grid normalisation) and made the paste header pre-flight O(names) instead of O(columns x names). No public API or behaviour change.

[0.1.35] — 2026-06-22

Changed

• Autopaste now separates a new Benchling table from any content above it. initialise_table_benchling(mode="autopaste") previously applied the table_gap separation only when stacking below another managed Benchling table (one carrying a *_table_start bookmark). A table placed below a manually built table or stray notes got zero blank rows, so its merged title row stuck directly to the content above. table_gap blank rows are now inserted below manual tables, notes, and managed tables alike.
  • An empty sheet still starts at row 1 (no leading gap).
  • Pass table_gap=0 to restore the previous flush placement.

[0.1.3] — 2026-06-08

Changed

• initialise_table_benchling now takes a schema_map dict ({sheet_name: [schema_id, ...]}) and returns a typed list[SchemaInitResult] (each with a tuple of ColumnInfo) instead of a list[dict]. Result fields are accessed by attribute (e.g. result.action, column.range_name).

Added

• Per-column named ranges are namespaced as {anchor}__{token} so multiple schemas can share a sheet without name collisions; readers still reference columns by their bare token.

[0.1.0] — [0.1.2]

• Initial PyPI releases: the package-first xlwings + pandas helper API (named-range tables, plate functions, Benchling table initialisation, formatting, plotting, and file I/O).