[PATCH] Refactoring of class hierarchy around JosmComboBox / AutoCompComboBox

[marcello@…]

Added classes in bold to orthogonalize hierarchy:

JComboBox	JComboBoxModel	JComboBoxEditor	JTextField
JosmComboBox	JosmComboBoxModel	JosmComboBoxEditor	JosmTextField
AutoCompComboBox	AutoCompComboBoxModel	AutoCompComboBoxEditor	AutoCompTextField

Code reuse: JosmComboBox now uses a JosmTextField as editor
Code reuse: AutoCompComboBox now uses AutoCompTextField as editor
JosmComboBox uses more of the original L&F
JosmComboBox lists now expand all the way to the bottom or the top of the screen
Variable height items in combobox lists now work #19321
Autocomplete uses different algorithm #21290
editable="false" comboboxes in Presets now work #6157 #11024 #18714
The user may toggle LTR-RTL script in JosmTextField (menu and ctrl+space)
LTR-RTL automatically toggles according to key in AddTag and EditTag dialogs #16163

[Don-vip]

Wow, that's quite a patch

[Don-vip]

Does it mean your patch solves #21290, #6157 and #16163 all at once?

[marcello@…]

It fixes #6157, proposes a solution for #16163, and hopefully fixes #21290 (lacking a Mac I cannot test it).

[Don-vip]

excellent!

[Don-vip]

In 18221/josm:

fix #21319 - Refactoring of class hierarchy around JosmComboBox / AutoCompComboBox (patch by marcello):

• Code reuse: JosmComboBox now uses a JosmTextField as editor
• Code reuse: AutoCompComboBox now uses AutoCompTextField as editor
• JosmComboBox uses more of the original L&F
• JosmComboBox lists now expand all the way to the bottom or the top of the screen
• Variable height items in combobox lists now work, see #19321
• Autocomplete uses different algorithm, fix #21290
• editable="false" comboboxes in Presets now work, fix #6157 see #11024 see #18714
• The user may toggle LTR-RTL script in JosmTextField (menu and ctrl+space)
• LTR-RTL automatically toggles according to key in AddTag and EditTag dialogs, fix #16163

[Don-vip]

In 18222/josm:

see #21319 - fix NPE

[skyper]

missing text

[skyper]

I have some missing <optional> with or without additional text which only appear when I open the preset for the second time. The first time opening the preset after restart/reload empty space is present but the text is missing. Is this related to this ticket?

missing	working

[skyper]

Just tried with r18218 which works as expected, so I blame r18221.

[marcello@…]

Oops. Sorry. Patch.

[Don-vip]

In 18223/josm:

fix #21319 - fix optional regression (patch by marcello)

[skyper]

Does this patch also fix <optional text="Hallo world">?

[marcello@…]

where do you see that?

[skyper]

<optional> has the option to add text instead of the default with text="".

In my screenshots above it is <optional text="Only in case of manual actuator or manual mechanical drive: "> and one of many examples in current defaultpresets.xml is <optional text="Edit Highway Attributes:"> in "Cycle Lane/Track" preset.

[skyper]

Just tried r18224 and it works. Thanks for the quick fix.

[gaben]

Hmm, combo boxes don't seem to work when a value has already entered. You can check it eg. in the restaurant preset.

[gaben]

Ah sorry, multiselect is broken, not combobox. r18218 works, r18222 not.

[skyper]

Replying to gaben:

Ah sorry, multiselect is broken, not combobox. r18218 works, r18222 not.

Cannot reproduce with r18237 and flatlaf (Dracula).

[gaben]

For me, both Win and Linux affected with default LaF, r18237. @skyper, can you please test with native laf?

[gaben]

Oh okay, so the multiselect is only "broken" when the tag has only a single value. Also if the value contains extra space between the semicolon separators eg. cuisine=asian; african.

By broken I mean greyed out, so I can not use the GUI for adding items.

[skyper]

Ok, I can reproduce it if an unknown value is present (including leading or trailing white spaces). This is nothing new and happens with r18193, too. The problem is that, atm, unlike combo no values are add to the list and cause of that not all values are displayed. To stop users from overwriting valid but unknown values the multiselect is disabled. Please, open a new enhancement ticket, if you think this should be changed. Thanks.

[gaben]

Created a new ticket, see #21385.

[skyper]

Another possible regression: #21397

[Don-vip]

Milestone renamed

[GerdP]

Another possible regression: #21438

[GerdP]

another regression? #21619

[taylor.smock]

Another regression: #22055

[taylor.smock]

In 18538/josm:

JosmTextField: setHint now properly returns the old hint

See #21319/r18221. This additionally adds a non-regression test.

[Klumbumbus]

There are 62 unit tests failing on Jenkins since r18538. Not sure if this is a problem on Jenkins though, since there are also 2 test failing after r18537, which doens't make sense imo.

[taylor.smock]

Replying to Klumbumbus:

There are 62 unit tests failing on Jenkins since r18538. Not sure if this is a problem on Jenkins though, since there are also 2 test failing after r18537, which doens't make sense imo.

The OSM dev server was down (errol). So all the integration tests expecting a live dev api failed.

[Klumbumbus]

Ah, OK. I just started a new run on Jenkins. Lets see if the tests pass this time. --> they did.

[taylor.smock]

Replying to Klumbumbus:

Ah, OK. I just started a new run on Jenkins. Lets see if the tests pass this time. --> they did.

That is kind of what I expected. I don't know how prevalent the dev server going down is going to be (this is the second or third time this year). There was an outage in July as well. I tracked that one in #22216.