[
https://issues.apache.org/jira/browse/WW-5811?focusedWorklogId=1050047&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-1050047
]
ASF GitHub Bot logged work on WW-5811:
--------------------------------------
Author: ASF GitHub Bot
Created on: 25/Oct/26 16:13
Start Date: 25/Oct/26 16:13
Worklog Time Spent: 10m
Work Description: lukaszlenart commented on code in PR #2061:
URL: https://github.com/apache/struts/pull/2061#discussion_r4371474506
##########
core/src/main/java/org/apache/struts2/components/Form.java:
##########
@@ -428,6 +438,51 @@ private List<Object> resolveChain(String path) {
return chain;
}
+ /**
+ * The declared type of the property a field binds to, found on the object
parameter binding would
+ * write to and then walked through declared types rather than values, so
a null intermediate object
+ * (which binding creates) does not hide it. Null for an indexed or
unknown path.
+ */
+ Class<?> getPropertyType(String name) {
+ if (name == null || name.isEmpty() || StringUtils.containsAny(name,
'[', '(')) {
+ return null;
+ }
+ String[] segments = name.split("\\.");
+ for (Object root : bindingRoots()) {
+ Class<?> type = declaredType(root.getClass(), segments[0]);
+ if (type != null) {
+ for (int i = 1; i < segments.length && type != null; i++) {
+ type = declaredType(type, segments[i]);
+ }
+ return type;
+ }
+ }
+ return null;
+ }
+
+ private List<Object> bindingRoots() {
+ Object invoked = currentAction();
Review Comment:
Fixed in 9da0b93d9. The type is now read on the form's submission action
class. When a different action renders the form, the declared types of the
submission class are used; for a `ModelDriven` submission class the model type
is only known at runtime, so no type is passed and no `required` is emitted.
Covered by
`testThePropertyTypeIsReadFromTheSubmissionActionWhenAnotherActionRendersTheForm`
and
`testAModelDrivenSubmissionActionRenderedByAnotherActionHandsTheProviderNoType`.
##########
core/src/main/java/org/apache/struts2/components/StrutsHtmlConstraintProvider.java:
##########
@@ -119,7 +137,11 @@ protected void addConstraints(Map<String, String>
attributes, Validator validato
} else if (validator instanceof DoubleRangeFieldValidator
doubleValidator) {
addDoubleRange(attributes, doubleValidator, control);
} else if (validator instanceof RangeValidatorSupport<?>
rangeValidator) {
- addRange(attributes, rangeValidator, control);
+ if (control == HtmlControlType.DATE) {
+ addDateRange(attributes, rangeValidator);
+ } else {
+ addRange(attributes, rangeValidator, control);
Review Comment:
Fixed in 9da0b93d9. `addConstraints` dispatches every range validator to
`addRange` again; `addRange` hands `DATE` to `addDateRange`, so an overriding
subclass keeps receiving it. Covered by
`aSubclassOverridingAddRangeStillHandlesTheDateControl`.
##########
core/src/main/java/org/apache/struts2/components/StrutsHtmlConstraintProvider.java:
##########
@@ -155,6 +178,63 @@ protected void addRequiredField(Map<String, String>
attributes, RequiredFieldVal
}
}
+ /**
+ * An empty temporal input submits {@code ""}, which the built-in
converters turn into null for the
+ * types {@link #convertsBlankToNull(Class)} accepts, so {@code required}
then fails on both sides.
+ *
+ * @since 7.5.0
+ */
+ protected void addTemporalRequired(Map<String, String> attributes,
HtmlControlType control, Class<?> propertyType) {
+ if (control.isTemporal() && convertsBlankToNull(propertyType)) {
+ attributes.put(REQUIRED_ATTRIBUTE, REQUIRED_ATTRIBUTE);
+ }
+ }
+
+ /**
+ * A property with its own converter that maps {@code ""} to a value needs
a provider overriding this.
+ *
+ * @since 7.5.0
+ */
+ protected boolean convertsBlankToNull(Class<?> propertyType) {
+ if (propertyType == null) {
+ return false;
+ }
+ return Date.class.isAssignableFrom(propertyType)
+ || Calendar.class.isAssignableFrom(propertyType)
+ || LocalDate.class.isAssignableFrom(propertyType)
+ || LocalDateTime.class.isAssignableFrom(propertyType)
+ || LocalTime.class.isAssignableFrom(propertyType)
+ || OffsetDateTime.class.isAssignableFrom(propertyType);
+ }
+
+ /**
+ * A {@code java.util.Date} bound becomes its calendar date in the default
time zone, the zone the
+ * submitted {@code yyyy-MM-dd} is parsed in. Dropping the time leaves the
browser no stricter than the
+ * server: a min of 10:00 on the 1st still lets the 1st through
client-side.
+ *
+ * @since 7.5.0
+ */
+ protected void addDateRange(Map<String, String> attributes,
RangeValidatorSupport<?> validator) {
+ putIfPresent(attributes, "min", toHtmlDate(validator.getMin()));
Review Comment:
Fixed in 9da0b93d9, in `UIBean` rather than the provider, since only the tag
knows its attributes: when the field sets its own `step`, the derived `min` is
dropped and `max` kept. This also covers numeric controls, where an integral
derived `min` had the same effect on a custom `step` since 7.4.0. Covered by
`testADeveloperStepSuppressesTheDerivedMinButKeepsTheMax`.
##########
core/src/main/java/org/apache/struts2/components/StrutsHtmlConstraintProvider.java:
##########
@@ -155,6 +178,63 @@ protected void addRequiredField(Map<String, String>
attributes, RequiredFieldVal
}
}
+ /**
+ * An empty temporal input submits {@code ""}, which the built-in
converters turn into null for the
+ * types {@link #convertsBlankToNull(Class)} accepts, so {@code required}
then fails on both sides.
+ *
+ * @since 7.5.0
+ */
+ protected void addTemporalRequired(Map<String, String> attributes,
HtmlControlType control, Class<?> propertyType) {
+ if (control.isTemporal() && convertsBlankToNull(propertyType)) {
+ attributes.put(REQUIRED_ATTRIBUTE, REQUIRED_ATTRIBUTE);
+ }
+ }
+
+ /**
+ * A property with its own converter that maps {@code ""} to a value needs
a provider overriding this.
+ *
+ * @since 7.5.0
+ */
+ protected boolean convertsBlankToNull(Class<?> propertyType) {
+ if (propertyType == null) {
+ return false;
+ }
+ return Date.class.isAssignableFrom(propertyType)
+ || Calendar.class.isAssignableFrom(propertyType)
+ || LocalDate.class.isAssignableFrom(propertyType)
+ || LocalDateTime.class.isAssignableFrom(propertyType)
+ || LocalTime.class.isAssignableFrom(propertyType)
+ || OffsetDateTime.class.isAssignableFrom(propertyType);
+ }
+
+ /**
+ * A {@code java.util.Date} bound becomes its calendar date in the default
time zone, the zone the
+ * submitted {@code yyyy-MM-dd} is parsed in. Dropping the time leaves the
browser no stricter than the
+ * server: a min of 10:00 on the 1st still lets the 1st through
client-side.
+ *
+ * @since 7.5.0
+ */
+ protected void addDateRange(Map<String, String> attributes,
RangeValidatorSupport<?> validator) {
+ putIfPresent(attributes, "min", toHtmlDate(validator.getMin()));
+ putIfPresent(attributes, "max", toHtmlDate(validator.getMax()));
+ }
+
+ private String toHtmlDate(Object bound) {
+ LocalDate date;
+ if (bound instanceof LocalDate localDate) {
+ date = localDate;
+ } else if (bound instanceof Date utilDate) {
+ date =
Instant.ofEpochMilli(utilDate.getTime()).atZone(ZoneId.systemDefault()).toLocalDate();
Review Comment:
Reproduced (`min="1500-01-10"` in UTC) and fixed in 9da0b93d9 by omitting
`java.util.Date` bounds before the Gregorian cutover rather than emulating the
hybrid calendar — an omitted bound only makes the browser more permissive.
Covered by
`dateRangeOmitsABoundBeforeTheGregorianCutoverWhereTheConverterReadsJulianDates`.
Issue Time Tracking
-------------------
Worklog Id: (was: 1050047)
Time Spent: 1h (was: 50m)
> Derive HTML5 min/max and required constraints for temporal controls
> -------------------------------------------------------------------
>
> Key: WW-5811
> URL: https://issues.apache.org/jira/browse/WW-5811
> Project: Struts 2
> Issue Type: Improvement
> Components: Core Tags
> Reporter: Lukasz Lenart
> Assignee: Lukasz Lenart
> Priority: Major
> Fix For: 7.5.0
>
> Time Spent: 1h
> Remaining Estimate: 0h
>
> h2. Problem
> HTML5 constraint derivation ({{struts.ui.html5.constraints}}) maps range
> validators to {{min}}/{{max}} only for {{number}} and {{range}} controls.
> {{StrutsHtmlConstraintProvider}} deliberately skips the temporal controls, so
> {{DateRangeFieldValidator}} (and the new {{localdate}} validator, WW-5816)
> emit nothing.
> The {{required}} validator has a related gap. {{addRequiredField}} emits
> {{required}} only for {{radio}} and {{file}}, because other controls submit
> an empty string, which the {{required}} validator accepts. A temporal control
> bound to a {{java.util.Date}} (or {{java.time}}) property is different:
> {{DateConverter}} turns {{""}} into null, so the server rejects the empty
> value just as the browser would. These controls currently get only
> {{data-msg-required}}.
> h2. Scope
> * {{min}}/{{max}} on *{{date}} controls only*, from {{LocalDate}} bounds
> ({{localdate}}) and {{java.util.Date}} bounds ({{date}}), written as
> {{yyyy-MM-dd}}. A {{java.util.Date}} bound becomes its calendar date in the
> server default time zone (the zone the submitted value is parsed in);
> dropping the time never makes the browser stricter than the server. Years
> outside 0001-9999 are skipped.
> * {{required}} on *all temporal controls* ({{date}}, {{month}}, {{week}},
> {{time}}, {{datetime-local}}) from a {{required}} validator, when the bound
> property's declared type converts {{""}} to null: {{java.util.Date}} and
> subclasses, {{Calendar}}, {{LocalDate}}, {{LocalDateTime}}, {{LocalTime}},
> {{OffsetDateTime}}. A {{String}} property gets no {{required}}.
> * The provider needs the property type, so {{HtmlConstraintProvider}} gains a
> {{default}} {{constraintsFor}} overload carrying it, which delegates to the
> 7.4.0 method.
> h2. Out of scope
> {{min}}/{{max}} on {{month}}, {{week}}, {{time}} and {{datetime-local}}: no
> built-in converter and range validator pair reads their submitted values
> ({{datetime-local}} submits without seconds, there is no
> {{LocalTime}}/{{LocalDateTime}} range validator, and {{month}}/{{week}} have
> no converter). A separate ticket if needed.
> Related: WW-5778 (datetextfield in the html5 theme), WW-5810 (java.time
> support in datetextfield), WW-5816 ({{localdate}} validator)
> PR: [apache/struts#2061|https://github.com/apache/struts/pull/2061], docs:
> [apache/struts-site#367|https://github.com/apache/struts-site/pull/367]
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
