Always relies on floatingLabelStyle when FloatingLabelBehavior.always

[bleroux]

Description

With this PR, when InputDecorator.floatingLabelBehavior is set to FloatingLabelBehavior.always the label style is always set to InputDecorator.floatingLabelStyle, previously InputDecorator.labelStyle was used when the field was not focused or was empty.

Related Issue

Fixes #147231

Tests

Adds 1 test for this particular issue and several missing tests.

[Renzo-Olivares]

I wonder if _labelShouldWithdraw should just take into account decoration.floatingLabelBehavior == FloatingLabelBehavior.always. I looked through the usages in this file and it is almost always used along with decoration.floatingLabelBehavior == FloatingLabelBehavior.always so it seems reasonable. What do you think?

[bleroux]

This was my first thinking.
Then I saw the _labelShouldWithdraw usage related to Semantics and prefix/suffix and I was not sure of the consequences.
See

flutter/packages/flutter/lib/src/material/input_decorator.dart

Lines 2176 to 2200 in 5f0628d

	final bool needsSemanticsSortOrder = widget._labelShouldWithdraw && (input != null ? (hasPrefix || hasSuffix) : (hasPrefix && hasSuffix));
	final Widget? prefix = hasPrefix
	? _AffixText(
	labelIsFloating: widget._labelShouldWithdraw,
	text: decoration.prefixText,
	style: MaterialStateProperty.resolveAs(decoration.prefixStyle, materialState) ?? hintStyle,
	semanticsSortKey: needsSemanticsSortOrder ? _kPrefixSemanticsSortOrder : null,
	semanticsTag: _kPrefixSemanticsTag,
	child: decoration.prefix,
	)
	: null;
	final Widget? suffix = hasSuffix
	? _AffixText(
	labelIsFloating: widget._labelShouldWithdraw,
	text: decoration.suffixText,
	style: MaterialStateProperty.resolveAs(decoration.suffixStyle, materialState) ?? hintStyle,
	semanticsSortKey: needsSemanticsSortOrder ? _kSuffixSemanticsSortOrder : null,
	semanticsTag: _kSuffixSemanticsTag,
	child: decoration.suffix,
	)
	: null;
	if (input != null && needsSemanticsSortOrder) {

But reconsidering this, moving the check to _labelShouldWithdraw makes sense and improve readibility because it simplifies several clauses.
I will push an update.

[Renzo-Olivares]

Yeah I think it makes sense to change this even in the semantics case, since its depending on whether the label is floating. I'll keep an eye out if this causes any issues.

[Renzo-Olivares]

Thank you for the contribution @bleroux, and for all the tests!

[Renzo-Olivares]

nit: I think you can get rid of the first or in this sentence.

[Renzo-Olivares]

LGTM w/ small nit. Thanks for the contribution!

[Renzo-Olivares]

I'll check on the google testing issues @bleroux.

[Renzo-Olivares]

@bleroux There is one test failure that i'm not too sure is expected or not. Before this change the prefix/prefixText would not show when the floatingLabelBehavior is set to FloatingLabelBehavior.always, and we unfocus the TextField. After this change the prefix/prefixText always shows when floatingLabelBehavior is set to FloatingLabelBehavior.always. Something similar also happens with suffix/suffixText. The main culprit is this block of code that builds the prefix and suffix

flutter/packages/flutter/lib/src/material/input_decorator.dart

Lines 2178 to 2198 in 3b2bfc7

	final Widget? prefix = hasPrefix
	? _AffixText(
	labelIsFloating: widget._labelShouldWithdraw,
	text: decoration.prefixText,
	style: MaterialStateProperty.resolveAs(decoration.prefixStyle, materialState) ?? hintStyle,
	semanticsSortKey: needsSemanticsSortOrder ? _kPrefixSemanticsSortOrder : null,
	semanticsTag: _kPrefixSemanticsTag,
	child: decoration.prefix,
	)
	: null;
	final Widget? suffix = hasSuffix
	? _AffixText(
	labelIsFloating: widget._labelShouldWithdraw,
	text: decoration.suffixText,
	style: MaterialStateProperty.resolveAs(decoration.suffixStyle, materialState) ?? hintStyle,
	semanticsSortKey: needsSemanticsSortOrder ? _kSuffixSemanticsSortOrder : null,
	semanticsTag: _kSuffixSemanticsTag,
	child: decoration.suffix,
	)
	: null;

. _AffixText depends on whether the label is floating to decide whether or not to visually show the prefix/suffix. Should the prefix/suffix always be displayed in this case? Let me know if this explanation was clear enough or if I can provide more context. I'm leaning towards the new behavior being the correct one, since I don't see a reason not to show the prefix if the label is floating.

[bleroux]

@Renzo-Olivares Many thanks for the detailed information.

I'm leaning towards the new behavior being the correct one, since I don't see a reason not to show the prefix if the label is floating.

I agree with you. I'm currently working on the M3 tests migration for prefix/suffix, I started looking closer at this behavior and I will post an update tomorrow to let you if I found anything that we missed.

[bleroux]

@Renzo-Olivares
I had a closer look and I still fully agree with you on "since I don't see a reason not to show the prefix if the label is floating."

I have updated the PR with two tests related to that behavior (and I will probably add more on the PR I'm working on which goals is to migrate M3 tests related to prefix/suffix).