When adding support for non-standard namespacing to Reader Xlsx, I changed most (hopefully all) the uses of string literals for the namespaces to class constants instead. Writer Xlsx naturally uses all the same namespaces, but has continued to use string literals. This PR replaces those with the same constants used by Reader Xlsx.
Almost all of these are handled through annotations. This shouldn't be our "go-to" solution, but it becomes necessary because Scrutinizer's analysis is often incorrect. Here is a typical example, from Cells.php.
```php
if ($this->currentCellIsDirty && isset($this->currentCoordinate, $this->currentCell)) {
$this->currentCell->detach();
```
Scrutinizer complains that `$this->currentCell` can be null here, but the `isset` condition guarantees that it must be non-null. Perhaps Scrutinizer is worried that `isset` might be overridden, and will accept only an explicit equality test for null for each of the isset arguments. Changing the code to do this seems riskier than just adding the annotation.
A far more common, and more frustrating, example is:
```php
foreach ($simpleXmlElement as $element) {
var_dump($element->method());
}
```
Scrutinizer complains that element might be null. I don't think it can. I have previously added code in places to eliminate the objection, and that may be a practical solution when `$element` is used many times in the loop. But, when it's used only once, annotating the objection away seems like a better solution (less overhead, clearer code). Many of the changes in this PR fall into this category.
* Fix Remainder of Calculation vs. Phpstan Issues
I had tried to include these changes as part of an earlier effort, but something about them broke Phpstan. I removed them until I could determine the actual cause, which is ...
The array `$phpSpreadsheetFunctions` is a very large and very complicated array, typehinted as `array`. It is declared as private static; however, its content never changes (at least not now - there are some outstanding proposals that might change that). I have had some success with changing unmodifiable private static to private const. However, such a change here causes Phpstan to perform a lot more processing and eventually time out. So, leave it as static.
Having identified the cause of the problem, none of the other changes were problematic, so this PR applies the rest of them.
* Scrutinizer
One problem.
* Scrutinizer False Positives Without Suggested Annotation
See if minor code changes can make these go away.
* Scrutinizer Still Experimenting
Trying again.
Fix#3095. Issue is not a problem for Php8, but is for 7.4. Code used to write Xml for float custom property uses locale-dependent (in Php7) cast. Change to use locale-independent (all Php releases) `sprintf('%F'...)` instead.
Fix#3093. PR #2595 added the ability for Xlsx Reader to accept a custom color palette. With no examples other than the one at hand, which included a full complement of 64 colors, that PR required 64 colors in the palette. It turns out that Mac Numbers exports to Excel with a palette with less than 64 colors. So, with an example of that at hand, relax the original restriction and accept a palette of any size.
Fix#1531. This is a replacement for PR #3081 (see last paragraph below), which I will close.
Calculation has a property `suppressFormulaErrors`, which really doesn't work as one might expect. If a calculation throws an exception, the setting of this property might prevent the Exception from being thrown, but it will still trigger an Error. I do not think this makes sense, and will change it so the calculation will return `false`, which is part of the original design but which would essentially never happen. This allows the user to save a corrupt spreadsheet, but this was already possible through the use of `setPreCalculateFormulas(false)` on the Writer, so this doesn't really open any new exposures. It nevertheless might be considered a breaking change because of the difference in behavior.
Deprecation - the visibility of the existing property is public, which means it can be changed directly. A new private property is added with a public setter/getter. The new property will be used when the existing property is null (default), which will allow the existing property to be deprecated.
Function getFunctions is changed to static - the array which it returns is static. Existing callers using it as non-static will still function correctly.
Although I am enabling this ability, I don't necessarily think it's a good idea to make use of it. See the original issue for a discussion of why. It is not mentioned in the official documentation, and I will not be adding documentation for it. The originator discovered it by reading the code, and I think that is sufficient for what will often be an ill-advised choice.
Many of the large number of problems with Calculation.php in phpstan baseline are addressed. PR 3081 ran afoul of something in phpstan. The changes in this ticket are more limited, adding a number of doc blocks but leaving executable code unchanged.
Scrutinizer had previously suggested annotations for 3 LookupRef tests, but it no longer accepts its own annotation for those cases. This PR cleans them up. ColumnsTest and RowsTest are extremely straightforward.
IndexTest is a bit more complicated, but only because, unlike the other two, it had no test which executed in the context of a spreadsheet. And, when I added those, I discovered a couple of bugs. INDEX always requires at least 2 parameters (row# is always required), but its entry in the function table specified 1-4 parameters, now changed to 2-4. And, omitting col# is not handled the same way as specifying 0 for col#, though the code had treated them identically. (The same would have been true for row# but, because it is now required, ...)
* Phpstan Baseline Fixes 2022-09-21
Eliminate about 200 more lines from Phpstan baseline. For Helper/Sample and Helper/Html and others, many properties are declared as protected despite the fact that the classes do not extend any other class, and there are no classes which extend them. They are changed to private; this could be a breaking change in circumstances for which I cannot think of a use case (user extends class for some reason).
* Slightly Botched Merge Commit
Hope this fixes phpstan.
* Xlsx Reader External Data Validations Flag Missing
Fix#2677. This PR supersedes #2679, written by @technghiath, which lacks tests, and probably doesn't solve the problem entirely. The code causing the problem appears to be the last remnant in Xlsx Reader which calls `children` using a namespace prefix rather than a namespace. That is changed, and tests are added where the tag is unexpectedly missing, and also where it uses a non-standard namespace prefix.
* Scrutinizer
Reports 1 "new" error. It isn't, but fix it anyhow.
* Fix One Existing Scrutinizer Problem
Only remaining problem in Reader/Xlsx.
* More Carets in composer.json
Following up on PR #3086, there are 3 additional items in the require-dev section that should have carets. I probably accidentally removed them for tcpdf and mitoteam, which point to the current release anyhow. Dependabot appears to be responsible for mpdf, which is not pointing to the current release, but I have tested with current successfully.
* Minor Fix for Change Made Earlier Today
"Scrutinizer Tweak" causes a Phpstan error. Fix baseline to correct it.
When I cloned this morning, composer gave me a message that the lock file was not up to date with the latest changes in composer.json. I do not understand why, but it suggested to run `composer update`, which I did. This led to a handful of problems with php-cs-fixer, all fixed with changes to doc-blocks, and phpstan (only Writer/Xls/Worksheet required a change to code). We would presumably have had these problems at the start of next month when dependabot did its thing, so fix them now.
Mark Baker
Rod Elias
oleibman
José Rodríguez
BrokenSourceCode