12. Migrating from 4.0 to 5.0 #

← Migrating from 3.0 to 4.0 · Contents · Release notes →

5.0 changes no signature of the thirty-seven components of 4.0, nor any event — except the button bar (buttonbar, 12.7) and the ue_link_clicked event of the Markdown viewer (markdownview), which receives a second argument, as_kind (12.5). Three components change words. The progress bar (progressbar) takes those of the gauges — ib_label, il_color and is_label_format disappear in favour of ib_show_value, il_value_color, is_format and is_unit (12.1). The cartesian chart (chartcartesian) writes its numbers through is_format, as the gauges and the pie chart do: of_set_value_format and of_clear_value_format disappear, and ii_point_count of a series becomes il_point_count (12.4). The crosstab (crosstab) takes the same road: of_set_value_format and of_clear_value_format disappear in favour of is_format — on the field, or on one value alone through of_value — and of ib_negative_red (12.8). The compiler shows you every line to rework. The side bar (listbar) changes one event: ue_item_rclicked also receives the screen point of the click, al_x and al_y (12.9). The tiles box (tilesbox) renames one event: ue_tile_context(as_keys) becomes ue_tile_rclicked(as_keys, al_x, al_y), and the position of of_move_tile and of_insert_tile becomes long al_position (12.12). The radial menu (radialmenu) loses four tooltip properties it never showed — is_tooltip, is_super_tooltip_title, is_super_tooltip_text and is_super_tooltip_image (12.5). The calendar (scheduler) changes three gesture events: ue_appointment_changing receives as_kind, ue_appointment_moved and ue_appointment_resized receive the values the appointment had (12.10). The shell explorer (shellexplorer) changes one event: ue_error also receives the path and a constant reason, as_path and as_reason (12.11). The rich text block (statictext) changes one event: ue_hyperlink also receives ab_opened, whether the link really opened (12.5). The tab (tab) changes one event: ue_tab_rclicked also receives the screen point of the click, al_x and al_y (12.13). The sound player (soundplayer) changes three events: ue_started, ue_ended and ue_failed also receive the number of the sound, al_sound (12.14). The toaster (toaster) changes one event: ue_toast_dismissed also receives the reason of the closing, as_reason (12.15). The web browser (webbrowser): of_go_back, of_go_forward and of_stop return a long, and under ib_veto_navigation an address typed by the user is asked about to ue_navigating (12.17). The toolbar (toolbar) changes one event: ue_item_rclicked also receives the screen point of the click, al_x and al_y (12.18). The video player (video) changes one signature: of_stop becomes a function returning a long (12.16) — a call written as a statement still compiles. Three components change a right-click event of an item, which also receives its screen point al_x, al_y: the button bar (ue_button_rclicked), the breadcrumb (breadcrumb, ue_item_rclicked) and the cartesian chart (ue_point_rclicked); the status bar (statusbar) gives that point like the others, no longer multiplied by the scale of the display (12.19). The foundation (12.20): the reason a non-visual component failed is read through of_get_last_error() — the is_last_error property disappears from crypto, restclient, soundplayer, speechout, xml, u_pbt_regex and u_pbt_zip —, and of_stop, of_close and of_reset of the non-visual components return a long.

The rest asks nothing of your code: the data grid (datagrid) and the ribbon (ribbon) gain numeric input masks, and the editor of a numeric column of the grid now opens on the text of the mask (12.2, 12.3). The new components of 5.0 are presented in the release notes.

12.1 progressbar: the words of the gauges #

5.0 brings two gauges, linear (gauge) and dial (angulargauge). A progress bar also shows a value on a scale: the library now speaks a single language for it, and the bar, which said the same thing in other words, takes the words of the gauges. No old name survives: neither a property nor a name accepted by of_set_property.

• is_format is a PowerBuilder display format applied to the value ("#,##0", "$#,##0.00"), with the separators of the display language, exactly as for the gauges. Empty, the bar shows the percentage, as before.
• The {percent}, {value}, {min} and {max} placeholders no longer exist. A label that mixed value and percentage ("{value} of {max} rows ({percent})") has no exact equivalent: choose the value (format and unit) or the percentage (and a unit). The maximum, which your code knows, is written in the unit.
• is_unit is new: a plain text written after the readout, separated by a space — after the value as after the percentage. of_reset clears it.
• A mask that holds % writes the share done, not the value: is_format = "0.0%" gives "35.0%" for 1,687 rows out of 4,820, rounded down like the percentage — the percentage with decimals that {percent} could not write.
• A paused or failed wait turns into a full bar. In 4.0, a busy bar (ib_indeterminate) given STATE_PAUSED or STATE_ERROR froze in its slide, often outside its track: the failure could not be seen. In 5.0 it stops as a full bar (a full ring) in the color of the state, like the Windows taskbar, and the screen reader no longer announces it busy but says its state.
• of_set_property("tooltip", …) and super_tooltip_* return -5: the bar publishes no tooltip, and the door by name, which set one, is closed like the properties.
• New, with nothing to rework: is_orientation (ORIENTATION_VERTICAL: a bar that stands up, in place of a VProgressBar), is_accessible_name (the name the screen reader says), and the text of a wait (is_unit alone while the bar is busy).

// 4.0 : the label wrote the value through placeholders
uo_progress.ib_label        = true
uo_progress.il_color        = RGB(/*red*/ 0, /*green*/ 120, /*blue*/ 215)
uo_progress.is_label_format = "{value} of {max} rows"

// 5.0 : the words of the gauges - a display format for the value, then a unit
uo_progress.ib_show_value  = true
uo_progress.il_value_color = RGB(/*red*/ 0, /*green*/ 120, /*blue*/ 215)
uo_progress.is_format      = "#,##0"
uo_progress.is_unit        = "of 500 rows"

The recipe: compile; every ib_label, il_color and is_label_format of a progress bar is reported. Replace the first two with ib_show_value and il_value_color. For the third: an empty text stays empty in is_format; a label that wrote the value becomes a format ("#,##0") and a unit; a label that wrote the percentage keeps is_format empty and puts its text in is_unit. An of_set_property that named label, label_format or color now returns -5 and does nothing (of_get_last_error names the property): search your code for these names.

12.2 datagrid: a numeric editor that types through a mask #

No signature changes, and ue_cell_edited still receives the bare number (1234.5). What changes shows on the screen: typing in a numeric column now follows an input mask, in the DataWindow syntax — the one you give through the new is_mask property of the column, otherwise its display format stripped of its literals.

// 5.0 : the input mask of a numeric column - the cell still receives the bare number
uo_grid.of_column(/*key*/ "revenue").is_mask = "#,##0.00"

The recipe: nothing to change. Code that read the typed text never sees it: ue_cell_edited still gives the number. A column whose input must stay free (a code, a part number) is a text column, not a numeric one.

12.3 ribbon: the masks of a spinner #

Nothing changes for a spinner without a mask. Two new properties of its handle, in the PowerBuilder syntax: is_mask, the input mask ("#,##0.00", "0%"), and is_format, the display format with its literals (0 "pt" shows 11 pt, typed 11). id_value and ue_value_changed keep the bare number, the swap between the two raises no event, and the masks follow the spinner into the panel of a collapsed group.

// 5.0 : a font size shown "11 pt", typed "11"
uo_ribbon.of_item(/*keys*/ "home/font/size").is_format = "0 ~"pt~""

The recipe: nothing to change.

12.4 chartcartesian: the number format through is_format #

The cartesian chart wrote its numbers through a function with four arguments (decimals, prefix, suffix, thousands separator). 5.0 puts the charts and the gauges under one language: a PowerBuilder display format, the one of a DataWindow column, in the is_format property — of the whole chart, and of an axis through its handle. The sections positive;negative;zero, the literals and the percentage come with it, and the separators stay those of the display language. No old name survives.

• The format of an axis wins over the chart's, and keeps it. Setting the format of the chart no longer clears that of the axes (the form of of_set_value_format without an axis did): an axis goes back to the chart's format with is_format = "".
• Without a format (empty, or a format that is not a numeric one), the number is written as it is, to the thousandth, with the decimal separator of the display language — as the default format did.

// 4.0 : decimals, prefix, suffix and thousands as four arguments
uo_chart.of_set_value_format(/*decimals*/ 0, /*prefix*/ "$", /*suffix*/ "K", /*thousands*/ true)
uo_chart.of_set_value_format(/*axis*/ u_pbt_chartcartesian.AXIS_VALUE2, /*decimals*/ 1, /*prefix*/ "", /*suffix*/ " u", /*thousands*/ false)

// 5.0 : a display format, written as in a DataWindow - for the chart, then for one axis
uo_chart.is_format = "$#,##0K"
uo_chart.of_axis(/*axis*/ u_pbt_chartcartesian.AXIS_VALUE2).is_format = '0.0" u"'

The recipe: compile; every of_set_value_format and of_clear_value_format is reported. Write the format as in a DataWindow: decimals gives the number of 0 after the point (-1 = #,##0.###), thousands the grouping ,, the prefix and the suffix are written before and after (in quotes when they hold digits or spaces). The format of an axis goes into of_axis(<axis>).is_format.

12.5 Other behaviours you may see #

The recipe: nothing to change.

12.6 What was added #

None of the following asks you to change your code. The eight new components of 5.0 are presented in the release notes.

12.7 buttonbar: key, label, then icon; the "&" and the chord #

The button bar spoke differently from the rest of the library: the icon before the label, the position as the second argument, and an is_shortcut that meant "Alt letter" where it means "chord" everywhere else — is_shortcut = "Ctrl+S" armed Alt+C. 5.0 aligns it with button, the toolbar and the PowerBuilder CommandButton.

• is_shortcut of a button is a chord. During 5.0, a single letter is still read there as the Alt letter of 4.0 (it goes to is_mnemonic); write the "&" in the label instead. A chord the window can never fire (a letter alone, Ctrl+Shift) is refused: is_shortcut reads back empty, of_register_shortcut returns -5.
• A greyed button no longer steals a key. The greyed default button of a bar, or the Alt letter of a greyed button, leaves Enter and the letter to the window — they no longer go to the button of ANOTHER bar (two MDI sheets); the chord of a greyed button leaves its key to the window (a menu accelerator, for instance). The arrow keys skip a greyed button.

// 4.0 : the icon before the label
uo_bar.of_add_button(/*key*/ "save", /*image_file*/ "img\save.png", /*text*/ "Save")

// The Alt letter went to is_shortcut (the letter S alone), the chord to of_register_shortcut
uo_bar.of_register_shortcut(/*chord*/ "Ctrl+S", /*key*/ "save")

// 5.0 : key, label, icon - the "&" of the label gives Alt+S
uo_bar.of_add_button(/*key*/ "save", /*text*/ "&Save", /*image*/ "img\save.png")

// The chord is a property of the button, shown in its tooltip
uo_bar.of_button(/*key*/ "save").is_shortcut = "Ctrl+S"

The recipe: compile; a call of of_insert_button is reported (the position goes last). A three-argument of_add_button still compiles: swap the last two (the label, then the icon) — a button whose label is a file path shows you the ones left. Replace every one-letter is_shortcut with a "&" in the label, and double every "&" that must be shown (R&&D).

12.8 crosstab: the format of the values through is_format #

The crosstab wrote the numbers of a measure through a function with four or five arguments (decimals, thousands separator, symbol, symbol before), which held for every value of a field: the same field twice in Values — the amount, and its share of the total — could not have two formats. 5.0 puts it under the language of the charts and the gauges: a PowerBuilder display format in is_format, on the handle of the field (of_field) for all its values, or on the handle of one value (of_value(position), new) for it alone. Empty, a value takes the Format of the DataWindow column. No old name survives.

• The thousands separator of a value with no format stays is_thousands, like the Options menu of the grid: that is what an of_set_value_format with an empty separator did. The currency the chip menu offers stays is_currency_symbol.
• A value with no format now shows the Format of its column of the DataWindow, read by of_from_datastore: an amount whose column carries #,##0.00 is shown with two decimals without a line of code.

// 4.0 : decimals, thousands, symbol and its side as arguments - for every value of the field
uo_cross.of_set_value_format(/*field*/ "amount", /*decimals*/ 2, /*thousands*/ "locale", /*suffix*/ "$", /*symbol_before*/ true)

// 5.0 : a display format, written as in a DataWindow - for the field, then for its second value alone
uo_cross.of_field(/*field*/ "amount").is_format = "$#,##0.00"
uo_cross.of_value(/*position*/ 2).is_format = "#,##0"

The recipe: compile; every of_set_value_format and of_clear_value_format is reported. Write the format as in a DataWindow: ai_decimals gives the number of 0 after the point, a thousands separator gives the grouping , ("none": none), the symbol is written before or after (in quotes when it holds digits or spaces). The format of one value alone goes into of_value(<position>).is_format.

12.9 listbar: the right click carries its point, a layout is announced #

The side bar joins the library on three points: the right click of an entry gives its screen point like the one of the trees, the grids and the kanban; an order of the code that changes the layout announces it like a gesture; a greyed entry no longer becomes the selection. It also gains an event for the click on the entry already selected, a search and the building from a DataStore (12.6).

// 4.0 : PointerX() of the userobject - the menu opened too high under a ribbon
m_entry.PopMenu(PointerX(), PointerY())

// 5.0 : the pointer of the WINDOW for a PowerBuilder menu (al_x, al_y serve a menu of the library)
m_entry.PopMenu(parent.PointerX(), parent.PointerY())

The recipe: nothing refuses to compile — a script of ue_item_rclicked keeps as_keys. Open your menu at the pointer of the window (parent.PointerX()), or hand al_x, al_y to of_show of a contextmenu. Set the initial selection before switching ib_veto_selection on, and make sure your ue_layout_changed accepts the call of_set_layout now makes to it.

12.10 scheduler: the kind of a change and the values it had #

The calendar joins the Gantt chart: the question asked before a gesture says whether it is a move or a resize, and the events that follow give the values the appointment had — a log, an undo or a confirmation no longer keep their own copy of every appointment. A selection your code takes away is announced, and four calls now refuse what they accepted in silence.

// 5.0 : ue_appointment_changing event of uo_sched : (string as_key, string as_kind, string as_start, string as_end, string as_resource) returns boolean
// A resize keeps the day : only a move is checked
if as_kind = u_pbt_scheduler.CHANGE_RESIZE then return true

// Nothing on a Saturday or a Sunday : returning false puts the appointment back
if DayNumber(Date(Left(as_start, 10))) = 1 or DayNumber(Date(Left(as_start, 10))) = 7 then return false
return true

The recipe: nothing refuses to compile — a script of ue_appointment_changing, ue_appointment_moved or ue_appointment_resized keeps its arguments, the new ones are added. If your ue_selection_changed enables buttons, it now receives "" when the selected appointment leaves; read again the return codes of of_add_appointment, of_scroll_to_time, of_set_field and of_show_appointment, which may now return -5. To save the gestures without code, see ib_write_back.

12.11 shellexplorer: the error says its reason #

The shell explorer now gives, in ue_error as in ue_path_not_found, a reason your code can test — a REASON_* constant of u_pbt_shellexplorer, the same in every language of Windows. It no longer raises ue_selected for the folder already selected, and announces with ue_collapsed a branch it closes by itself. It also gains the right click on a folder (ue_node_rclicked) and the address of the root.

// 5.0 : ue_error event of uo_tree : (string as_path, string as_reason, string as_message)
// Test the reason, never the text of the shell
if as_reason = u_pbt_shellexplorer.REASON_TIMEOUT then
    st_status.text = "The network folder " + as_path + " does not answer"
else
    st_status.text = as_message
end if

The recipe: nothing refuses to compile — a script of ue_error keeps as_message, the new arguments are added. A test on the TEXT of as_reason in ue_path_not_found ("not found", "outside the root") is written again with a REASON_* constant; an application that reloaded a pane at every ue_selected now does it once per change.

12.12 tilesbox: the right click carries its point, an order of the code is announced, the layout leaves the rights to the application #

The tiles box joins the library: its right click gives the screen point like the one of the trees, the grids and the kanban; an order of the code that moves a tile, folds a group or changes a size raises the same events as a gesture; the layout no longer carries the visibility, which stays the application's business. It also gains moving a tile into another group from the code (12.6).

// 4.0 : the right click of a tile
// ue_tile_context event of uo_tiles : (string as_keys)

// 5.0 : the same script, under the new name, with the screen point
// ue_tile_rclicked event of uo_tiles : (string as_keys, long al_x, long al_y)

The recipe: compile; your script of ue_tile_context stays on an event that no longer exists — move it to ue_tile_rclicked, which keeps as_keys. If your ue_layout_changed saves the layout, it now also sees what your code changes; a ue_tile_moved that must only react to the user ignores the moves your code has just made. If you kept the visibility in the layout, set it from the code at opening (ib_visible), according to the rights.

12.13 tab: the right click carries its point, an applied layout is announced #

The tab joins the library: the right click of a tab, built-in menu off, gives its screen point like the one of the trees, the grids and the kanban, and of_set_layout that changes the layout announces it like a gesture. A drag given back to its place no longer announces anything. It also gains pages created on demand (ue_page_needed, of_supply_page), the double click (ue_tab_double_clicked), ib_cycle_keys, is_close_mode, ii_strip_width and, on a page, ib_modified and is_badge — see the page of the tab.

// 4.0 : the right click of a tab, built-in menu off
// ue_tab_rclicked event of uo_tabs : (string as_key)

// 5.0 : the same script, with the screen point
// ue_tab_rclicked event of uo_tabs : (string as_key, long al_x, long al_y)

The recipe: nothing refuses to compile — a script of ue_tab_rclicked keeps as_key. Open your menu at the pointer of the window (parent.PointerX()), or hand al_x, al_y to of_show of a contextmenu. Make sure your ue_layout_changed accepts the call of_set_layout now makes to it.

12.14 soundplayer: the number of a sound, a refusal returned at once #

The sound player changes three things your code can see. Its events carry the number of the sound: ue_started, ue_ended and ue_failed receive a last argument, al_sound — the number of_play_over and of_enqueue now return (they returned 0), and that the new of_stop_sound takes. A source refused at once is a code: of_play, of_play_over, of_crossfade and of_preload return -5, the reason in of_get_last_error(), for an empty source, an http:// address, a malformed file:// address or a format refused by its name — and no longer raise ue_failed; the sound playing goes on. In the queue (of_enqueue), the refusal still comes through ue_failed, at its turn. Other -5: of_pause and of_resume with nothing to suspend or resume, of_beep and of_beep_sync at 0 Hz or 0 ms, of_meter out of 1 to 64 bars.

// Local variables
string ls_url
long ll_rc

// The address of the sound, typed by the user
ls_url = sle_url.text

// 4.0 : of_play returned 0, and an http:// address came back later through ue_failed
// 5.0 : the refusal is a code, at once ; of_get_last_error() gives the reason
ll_rc = inv_sound.of_play(/*source*/ ls_url)
if ll_rc = -5 then
	MessageBox("Sound", inv_sound.of_get_last_error())
end if

What changes without asking anything of your code: of_stop now dies out over il_fade_ms, as the documentation promised (ue_stopped still comes at once); when il_timeout_ms runs out, only the sound of the synchronous call is cut — the background, the overlays and the queue go on; of_crossfade to a refused source keeps the track playing; of_level is 0 and of_spectrum returns no bar, meter off; of_equalizer(ab_on, ai_gains[]) sets back to 0 a band the array does not reach; of_clear_queue also forgets the sounds the queue already played; the messages of ue_failed follow the display language (as_reason stays the word to test). The player also gains the queue that moves (of_next, of_previous, of_queue_index, ue_source_changed, ib_queue_repeat), the loop without a gap (ib_gapless_loop, REASON_TOO_LONG), the segment (of_play_range), the mute (ib_muted) and the buffering (ue_buffering, of_is_buffering) — see the page of the sound player.

The recipe: nothing refuses to compile — a script of ue_started, ue_ended or ue_failed keeps its arguments, al_sound is added. Read again the return-code tests of of_play_over and of_enqueue (a = 0 becomes > 0), and the code that waited for ue_failed for an http:// address or an empty source: it now tests the -5 of the call.

12.15 toaster: the reason of a closing, a toast closed by the code always says so #

The toaster says why a toast closed: ue_toast_dismissed receives a last argument, as_reason — REASON_TIMEOUT, REASON_USER (its cross), REASON_CODE (of_close, of_close_all) or REASON_OWNER_CLOSED (its window closed). And every toast closed by the code announces it, even closed right after of_show, before it was drawn, or while it waited for room: an application that waited for ue_toast_dismissed to free its state no longer depends on a race.

// 4.0
// ue_toast_dismissed event of inv_toaster : (long al_id, string as_key)

// 5.0 : the same script, with the reason
// ue_toast_dismissed event of inv_toaster : (long al_id, string as_key, string as_reason)

The recipe: nothing refuses to compile — a script of ue_toast_dismissed keeps al_id and as_key, as_reason is added. Read again the code that assumed a toast closed at once by of_close would say nothing: it now announces it. Toasts stay tied to the toaster that showed them: an of_close on the id of another toaster returns -5. The toaster also gains of_close_all (close all its toasts) and of_get_last_error — see the page of the toaster.

12.16 video: a refusal returned at once, a film and its subtitles in one call, the same film resumed #

The video player changes five things your code can see. A refusal is returned at once, as in the sound player (12.14): of_play(), of_play(source) and of_play_range return -5, the reason in of_get_last_error(), for an empty source, an http:// address or a format refused by its name (avi, wmv, flv) — and no longer raise ue_failed; the film playing goes on. The is_source property keeps its ue_failed: a property has no return code. The same film is not reloaded: of_play(source) or is_source on the film already loaded resumes it where it is (ue_resumed) instead of starting it over; of_seek(0) starts it over. A seek stays in the segment of of_play_range, the user's as well as of_seek: Home goes to the start of the segment, a click past its end to its end — before, Home left the extract and a click past it stopped the film. A playlist entry that cannot play raises ue_source_changed first, then one ue_failed — it came before, and twice. of_stop becomes a function: it returned nothing (a subroutine), it returns 0, or -2 when the component is not created — like of_stop of the web browser.

// Local variables
string ls_path
long ll_rc

// The film chosen by the user
ls_path = sle_film.text

// 4.0 : of_play returned 0, and an http:// address or an .avi came back later through ue_failed
// 5.0 : the refusal is a code, at once ; the reason is in of_get_last_error
ll_rc = uo_video.of_play(/*source*/ ls_path)
if ll_rc = -5 then
	MessageBox("Video", uo_video.of_get_last_error())
end if

What changes without asking anything of your code: a marker reached by a seek WHILE PAUSED no longer raises ue_marker_reached (only ue_seeked); Space and the Play button of a player with no film no longer do anything (they raised ue_failed(REASON_NO_SOURCE)); the subtitle language the user chose follows to the next film of the playlist; an http:// poster is refused and is_poster reads back empty; a refused is_subtitles reads back empty; a subtitle file of the workstation that cannot be found raises ue_subtitles_failed(REASON_NOT_FOUND); the speed menu works at the keyboard and its arrows no longer change the volume; in full screen the bar shows at the keyboard; the remembered position (ib_remember_position) no longer depends on how the path is written, and only the last 200 films are kept (a position of 4.0 is read once more). The player gains of_play(source, subtitles) (the film and ITS subtitles: set before a second film, they went with the first one), ib_forward_seek, of_watched_ms, of_watched_percent, ue_watched_completely and ib_skip_failed — see the page of the video player.

The recipe: nothing refuses to compile. Read again the code that waited for ue_failed after of_play for an http:// address, an empty source or an .avi: it now tests the -5 of the call. Read again too a "Play" button that relied on of_play(source) to start over: add of_seek(0).

12.17 webbrowser: a gesture of the user is asked about, a page of of_show_html has no address #

With ib_veto_navigation = true, ue_navigating is now also asked for a gesture of the user: an address typed in the bar, Open link of the context menu, a new window opened by a click. In 4.0 these three paths did not ask. Only an address set by your code is never asked about, and nor is a redirection of the server: for a list of servers that also holds for redirections, set the new property is_allowed_hosts.

The recipe: nothing refuses to compile, an ignored long compiles. A script of ue_navigating written for the links of the site now also sees the addresses typed by the user: if it decides on the address, let it compare its host, as the example of the component page does — or hand the list to is_allowed_hosts. The code that read as_url of ue_load_completed, or is_address, after of_show_html receives an empty string.

12.18 toolbar: the right click carries its point, an order of the code is announced #

The toolbar joins the library: the right click of a tool gives its screen point, like that of the trees, the grids and the tab, and moving a bar by ii_band / ii_index raises ue_bar_reordered like the gesture. Two adds refuse what they accepted silently, and a shortcut set where nothing lives is no longer registered. It also gains is_text on a bar, a name for every tool and field read by a screen reader, and a keyboard that crosses the fields — see the toolbar page.

// 4.0 : the right click of a tool
// ue_item_rclicked event of uo_toolbar : (string as_keys)

// 5.0 : the same script, with the screen point
// ue_item_rclicked event of uo_toolbar : (string as_keys, long al_x, long al_y)

The recipe: nothing refuses to compile — a script of ue_item_rclicked keeps as_keys. Open your menu at the point al_x, al_y (hand it to of_show of a contextmenu). Check that a script of ue_bar_reordered also accepts the moves your code makes, test the return code of of_add_datepicker when the date comes from a regional field, and set is_shortcut after the add of the tool or the entry.

12.19 The right click of an item carries its point: buttonbar, breadcrumb, chartcartesian, statusbar #

One rule for every right click on an item of the library: the event receives the key (or the address) of the item, then al_x, al_y, the point where to open a menu, in screen pixels — the pointer, or, from the keyboard (Menu key, Shift+F10), the bottom-left corner of the item (bottom-right in a right-to-left reading; the middle of a shape that is no rectangle: a point of a chart). A contextmenu of the library opens there as they are: uo_menu.of_show(al_x, al_y). Three events that did not carry it receive it; the status bar multiplied its own by the scale of the display, and a menu opened at that point landed far from the pointer on a display at 125 % or 150 %.

// 5.0 : the menu of a button, opened where the user right-clicked it
// ue_button_rclicked event of uo_buttons : (string as_key, long al_x, long al_y)
is_menu_button = as_key
uo_menu.of_show(/*x*/ al_x, /*y*/ al_y)

The recipe: nothing refuses to compile — a script of these events keeps its arguments. Open your menu at the point al_x, al_y (hand it to of_show of a contextmenu); a PowerBuilder menu keeps PopMenu(PointerX(), PointerY()) of your window. A script that corrected the scale of the coordinates of ue_panel_rclicked itself must no longer do it.

12.20 The foundation: a component's error is its own, every verb returns a code #

Five changes touch the whole library, or a whole family of components. Only one refuses to compile: is_last_error of a non-visual component, now of_get_last_error().

// Local variables
string ls_why

// 4.0 : ls_why = inv_crypto.is_last_error
// 5.0 : the reason of a failure, by the same method as every component
ls_why = inv_crypto.of_get_last_error()

The recipe: the compiler shows each .is_last_error to replace with .of_get_last_error(). A call to of_stop, of_close or of_reset written as a statement still compiles. A script that read of_get_last_error() of a component to learn the failure of another call (an export of the DLL called directly, another component) now reads that component's own reason, or nothing: ask the one that failed.

Added, nothing to rework: is_accessible_name and is_accessible_description on every visual component — the name and the description a screen reader says (3.4); of_from_datastore of a DataWindow or a DataWindowChild on crosstab and datagrid, as on the other data-bound components; the full form of of_insert_command (commandpalette), which takes the hint, the shortcut and the keywords as of_add_command does, with a long position.

← Migrating from 3.0 to 4.0 · Contents · Release notes →