RFR: 8234959: FXMLLoader does not populate ENGINE_SCOPE Bindings with FILENAME and ARGV

[Kevin Rushforth]

On Sun, 23 Feb 2020 14:57:44 GMT, Rony G. Flatscher <github.com+60214806+ronyfla at openjdk.org> wrote:

>> OK, forgot to submit an explanatory text related to this fix. Posted [1] which explains the problem and the suggested solution for discussion. 
>> 
>> - This pull request relates to: <https://bugs.openjdk.java.net/browse/JDK-8234959>
>> 
>> - Brief background: the controller code for a FXML file can be written in
>>   any of the Java script languages that implement the javax.script framework
>>   introduced with Java 6
>> 
>> There are three possible types of script code definitions possible in fxml files:
>> 
>> 1. in an empty script element where the source attribute denotes an external file
>>    that contains the code, e.g.:
>> 
>>      <fx:script source="somescript.rex" />
>> 
>>    This pull request uses the source attribute value as the filename.
>> 
>>    There are no arguments to be supplied.
>> 
>> 2. text of a script element, e.g.:
>> 
>>       <fx:script>say 'code executed at:' .dateTime~new </fx:script>
>> 
>>    or:
>> 
>>       <fx:script><![CDATA[say 'code executed at:' .dateTime~new ]]></fx:script>
>> 
>>    This pull request uses the fxml-filename appended with the string
>>    "-script_starting_at_line_" and the starting line number of the code.
>> 
>>    There are no arguments to be supplied.
>> 
>>    Reasonings:
>> 
>>         - in the case that multiple fxml files employ script code it is important
>>           to learn the name of the fxml file that hosts the code that gets executed
>> 
>>         - the dash after the fxml filename is meant to ease parsing
>> 
>>         - in the case that there are multiple fx:script elements in a fxml file
>>           with inline code it is important to learn in which line the code starts
>>           that gets executed
>> 
>>         - the decorated filename thereby becomes self documentary and unambiguous to
>>           the script code developer when debugging
>> 
>> 
>> 3. PCDATA text in an event attribute which by convention starts with "on", e.g. "onAction":
>> 
>>    <fx:button text="Hi!" onAction="say 'click at:' .dateTime~new', event='arg(1)~toString" />
>> 
>>    This pull request uses the fxml-filename appended with a dash "-", appended with the
>>    name of the event attribute, the string "_attribute_in_element_ending_at_line_" and
>>    the ending line number of the element.
>> 
>>    In the ScriptEventHandler constructor the decorated filename gets stored in addition
>>    to the script and the scriptEngine.
>> 
>>    Each time the event fires the event object argument will be stored in a copy of the
>>    engineBindings with the index EVENT_KEY as well as with a single capacity Object array
>>    stored with the index ScriptEngine.ARGV and the filename with the index
>>    ScriptEngine.FILENAME. The script's evaluation will get this engineBindings as
>>    its ENGINE_SCOPE Bindings supplied.
>> 
>>    Reasonings:
>> 
>>         - in the case that multiple fxml files employ script code it is important
>>           to learn the name of the fxml file that hosts the event code that gets
>>           executed
>> 
>>         - the dash after the fxml filename is meant to ease parsing
>> 
>>         - in the case that there are multiple event attributes in elements
>>           in a fxml file it is important to learn which event attribute in which
>>           element hosts the event script code that gets executed
>> 
>>         - the decorated filename thereby becomes self documentary and unambiguous to
>>           the script code developer when debugging
>> 
>>         - the structural change to the constructor (adding the decorated filename as
>>           a third argument) of the private static ScriptEventHandler class is confined
>>           to the FXMLLoader class and has no side effects
>> 
>> [1] "Ad suggested test unit for 'JDK-8234959 FXMLLoader does not populate ENGINE_SCOPE Bindings with FILENAME and ARGV'": <https://mail.openjdk.java.net/pipermail/openjfx-dev/2020-February/025104.html>
> 
> Sorry, mixed up the link with the test unit WIP! :(
> 
> This is the correct link with the explanatory text related to this suggested fix. Posted [1] which explains the problem and the suggested solution for discussion. 
> [1] 'Ad suggested fix for "JDK-8234959 FXMLLoader does not populate ENGINE_SCOPE Bindings with FILENAME and ARGV"': <https://mail.openjdk.java.net/pipermail/openjfx-dev/2020-February/025102.html>

@aghaisas can you also review this?

-------------

PR: https://git.openjdk.java.net/jfx/pull/122