feat: add drawing pass-through support for unsupported elements

[rpg600]

feat: add drawing pass-through support for unsupported elements

Add opt-in pass-through mechanism to preserve unsupported drawing
elements (shapes, textboxes etc) during read/write operations.

• Add enableDrawingPassThrough property to BaseReader
• Store original drawing XML when pass-through is enabled
• Reuse stored XML in Writer instead of regenerating
• Disabled by default for backward compatibility

This allows preservation of Excel elements not yet supported by
PhpSpreadsheet without requiring full implementation.

Currently, the pass-through mechanism works as follows:

1. Reader: setEnableDrawingPassThrough(true) stores the original drawing XML
2. Writer: Automatically uses the stored XML if:
   • The drawing collection is empty (no drawings added programmatically)
   • AND the unparsed data contains the original XML

I'm open to adding an explicit Writer flag if you believe it provides better clarity and control.

This addresses existing issue : #4037

[oleibman]

I am still evaluating whether this is a good idea or not. Will it work if you add or remove a row or a column on the sheet? Add a comment when there weren't any before, and when there were? Remove a comment?

Your commentary says that you are fixing 3 open issues. Are there tests for each of those?

[rpg600]

Will it work if you add or remove a row or a column on the sheet?

Yes but the drawing's coordinates (col, row, offsets) will not be adjusted.

Add a comment when there weren't any before, and when there were? Remove a comment?

Comments are completely independent from drawings.

Your commentary says that you are fixing 3 open issues. Are there tests for each of those?

I was too quick when referencing the issues, this PR actually only addresses one of them.

I added several unit tests, including checks for unsupported elements (shapes/textboxes), behavior without pass-through, comment deletion, row/column operations, and correct flag handling between the Reader and Writer.

This feature isn’t ideal, as it’s essentially a workaround, but it will provide a temporary solution while we wait for those implementations, and it will help avoid patching the code.

[oleibman]

Thank you for adding the tests.

Comments are, alas, not completely independent from drawings. It's why I asked the question. Create a new spreadsheet with a comment. Notice that there is a drawings folder with a vml file in it in the saved spreadsheet. I can't offhand tell whether this will be a problem for your change.

[rpg600]

I have added testCommentsAndPassThroughCoexist
It shows that VML drawings and DrawingML coexist in the xl/drawings/ folder without interference. Both existing and new comments are preserved correctly when pass-through is enabled.

[oleibman]

It is not absolutely required that Coveralls doesn't report a decrease. However, we strive to eliminate that situation if possible, especially in code which is new in a change. In Writer\Xlsx\Drawing.php, you have the following new statements which include some "misses" around line 617:

        if (!isset($unparsedLoadedData['sheets'][$codeName])) {
            return null;
        }

        $sheetData = $unparsedLoadedData['sheets'][$codeName];
        if (!is_array($sheetData)) {
            return null;
        }

        // Only use pass-through XML if the Reader flag was explicitly enabled
        if (($sheetData['drawingPassThroughEnabled'] ?? false) !== true) {
            return null;
        }

        if (!isset($sheetData['Drawings']) || !is_array($sheetData['Drawings'])) {
            return null;
        }

I tthink if you changed that block as follows, your code would still work as planned and there would no longer be any coverage misses:

        $sheetData = $unparsedLoadedData['sheets'][$codeName] ?? null;
        // Only use pass-through XML if the Reader flag was explicitly enabled
        if (!is_array($sheetData) || ($sheetData['drawingPassThroughEnabled'] ?? false) !== true || !is_array($sheetData['Drawings'] ?? null)) {
            return null;
        }

Please give that a shot.

[oleibman]

Thank you - success on both fronts - no problems with unit tests, and no missed statements.