See #1285, which went stale and was closed, but recently received some positive feeback. It seems easy to implement, and the only other plaintext file format, Html, allows loading from a string. Those two reasons combined suggest that we should do it.
Drawings in an Xlsx file are stored in such a way that Php can read their contents using the `zip:` protocol. This does not, however, work when the file is read by PhpSpreadsheet and then saved as Html or Pdf, since the browser will not recognize that protocol even if the file is available. Such drawings need to be saved in the html as embedded images in order for the copy to display them properly. This is true even when the writer is set to not embed images (default).
An additional problem arises when an Html file with an embedded image is read, because `Worksheet\Drawing::setPath` attempts to validate the path, which it cannot do for the `data:image` Url which embedded images use.
And yet another problem. Writer/Html writes out a MemoryDrawing as a png using the imagepng function; but then declares it as jpeg in the Html. This is now corrected.
And a fourth problem. Writer/Html ignores the last row if it contains nothing but a Memory Drawing, which can be true when copying an Xls file.
These changes are testable (it's how I discovered the second part of this parlay). I think it is also useful to add a sample to see the results of this type of copy.
* Real Errors Identified in Calculation by Scrutinizer
Before Scrutinizer broke, I took a look at the remaining 43 errors which it categorized as 'major'. Most of these were false positives, but, in the case of Calculation and Reader/Xlsx/Chart, I was able to determine that its analysis of some of the problems was correct. There is little point addressing the false positives until it starts working again, but we should fix the real errors.
This PR addresses the real errors in Calculation.
- A test for `$pCellParent` should have been a test for `$pCellWorksheet`.
- A test for `$operand1Data['reference']` should have been a test for `$operand1Data['value']`.
- A test for `$operand2Data['reference']` should have been a test for `$operand2Data['value']`.
* Fix Attempt to Erroneously Call trim on Array
Fix#2780. Warning message is being issued when getting calculated value of cell with value `=INDIRECT(ADDRESS(ROW(),COLUMN()-2))/$C$4`. This appears to be the case for all recent (and probably not so recent) releases; it is not a result of changes to the code base. Fix added to this PR because the erring section of code was proximate to code already changed in the PR. Test added.
* Minor Code Changes
Apply some suggestions from @MarkBaker
PHP 8.2 is supposed to deprecate the use of `['self', 'functionname']` for callables, suggesting the use of `[self::class, 'functionname']` instead. We made this change in a recent PR, and, while I'm thinking about it, I'll fix the remaining 2 modules with this construction. Vlookup is already adequately covered in unit tests. Reader/Xls/MD5 is not; a unit test is added.
Fixed that bug, using a slightly faster algorithm for the sort index than the simple fix would have used, and modified the tests that didn't have the correct expected result :-(
Fix#2768. DateFormatter handles only one of six special formats for time intervals `[h] [hh] [m] [mm] [s] [ss]`. This PR extends support to the rest. There should be no more than one of these in any format string. Although it certainly could make sense to treat `[d] [dd]` in the same manner, Excel does not seem to support those.
Interesting observations - hours and minutes are truncated (presumably because they may be followed by minutes and seconds), but seconds are rounded. Also, there are some floating point issues, which fortunately showed up for the example in the original issue. There, the time interval was 1.15, which should evaluate to a minutes value of 1656 (as it does in Excel). However, on my system it evaluated to 1655 because of a rounding error in the 13th decimal place. To overcome this, values are rounded to 10 decimal places before truncating.
* Fix reading of files in the root of a zip
Xlsx.php relies in dirname($filename) for path generation. When path is a bare filename (i.e. files in the root of the zip file), dirname($filename) returns a relative path to the current directory ("."). This is ok for filesystems, but not when accesing contents in a zip file.
Xlsx documents with files in the root of the zip container are not common, but legit. I've found it to happen in files generated by Google Campaign Manager 360.
* Update Xlsx.php
* Update Xlsx.php
* Update CHANGELOG.md
* Add files via upload
* Create XlsxRootZipFilesTest.php
* Update XlsxRootZipFilesTest.php
* Add files via upload
* Delete rootZipFiles.xlsx
* Update XlsxRootZipFilesTest.php
* Update Xlsx.php
PR #2720 failed because a timestamp in Document Properties Test was off by 1. This was due to one of two possible reasons. The constructor for Properties set the Created and Modified times using separate calls to the time function; if those happened to occur in different seconds, the test would fail. The test might also fail if the Created and Modified times used the same timestamp, but the time used to compare against those was calculated in a different second. It is surprising that this failure hasn't shown up before. Regardless, this PR corrects both possible problems.
It gets awkward when the defined name is for an actual range rather than for an individual named cell; because we need to manipulate the stack when that happens.
The code is ugly, and this is a rather simplistic approach, but it works as long as the named range is a cell, a cell range, or even a "chained" range - it won't work if we have union or intersection operators in the defined range - but it does provide formula support that never existed before.
Validation added for
- invalid characters
- invalid names ("C", "c", "R", or "r")
- cell references
- space separate words
- maxlength of 255 characters
- unique table names across worksheet
Current implementation for all methods that take a single cell reference argument:
- `setCellValue()`
- `setCellValueExplicit()`
- `getCell()`
- `cellExists()`
- `setBreak()`
- `freezePane()`
- `getComment()`
Also introducing a CellRange object to work with similar cases for methods that accept a cell range rather than simply a cell address; and RowRange/ColumnRange objects for those cases.
Still need to apply to methods that accept a cell range or single cell:
- `mergeCells()`
- `unmergeCells()`
- `protectCells()`
- `unprotectCells()`
- `setAutoFilter()`
Then there's a few special cases that accept row and column ranges, not simply cell ranges; or a series of cell ranges.
The code could stil do with some cleaning up, and better optimisation for memory usage; but all tests are passing... that's for full multi-level sorting (including direction), and allowing for correct sorting of sting/numeric datatypes.
* Eliminate Most Scrutinizer Problems in Test Suite
Mostly minor code changes, with some annotations.
* Missed 2 php-cs-fixer Problems
They should be fixed now.