FWD » Core Development » User Interface

Presumably since OpenEdge 11.3.0, "the PAUSE statement was enhanced for a whole or a fractional time-out interval in seconds. When the time-out period is a fractional value, it is rounded to the nearest whole millisecond.".

The similar behavior need to be implemented in FWD.

This is not hard to implement in FWD.

The PAUSE statement, READKEY and the WAIT-FOR statement all have this change. All of them should be updated accordingly to handle double instead of int. The link between these will be explained shortly.

PAUSE is implemented in static method calls logicalTerminal.pause() and LogicalTerminal.pauseBeforeHide(). Both of these already can accept a decimal and double type. These in turn call down to the following LogicalTerminal worker methods:

private static void pauseWorker(int seconds, String text, boolean noMessage, handle hWin)
public static void pauseBeforeHide(double seconds, String text, handle hWin)

WAIT-FOR is implemented as static method calls to LogicalTerminal.waitFor() which can handle decimal and double. It calls the LogicalTerminal worker:

public ScreenBuffer[] waitFor(EventList wc, int focusWidgetId, int seconds, ScreenBuffer[] inbuf)

READKEY is implemented as static method calls to KeyReader.readKey() which can take decimal and double. In turn, these call LogicalTerminal.readKey(int) or Stream.readChar(int pause).

All of these call to the FWD client for actual implementation, because there is a tight coupling of this feature with the UI. This is not just a simple sleep().

On the client side (see ClientExports and ThinClient) these are the API methods that get called:

public int pause(int seconds, String text, boolean noMessage, int windowId)
public void pauseBeforeHide(int seconds, String text, int windowId)
public int readKey(int seconds)
public ScreenBuffer[] waitFor(EventList wc, int focusWidgetId, int seconds, ScreenBuffer[] inbuf)

This change must flow through the downstream usage as well. In the ThinClient, the private Event waitForEvent(int seconds, boolean frameAsTarget, boolean honorServerEvent, boolean includeOSEvent) is used for all of the above (except for the Stream case). In turn, this depends on TypeAhead.getKeystroke(int) which takes a milliseconds value.

The actual waiting is only done in 2 places. If the double is passed through to ThinClient.waitForEvent() and Stream.readChar(int pause), I think the rest will work automatically.

No conversion changes are needed.

Greg Shah wrote:

• This use of PAUSE <milliseconds> is not currently supported. See #4472. The changes described there should be implemented.

Constantin wrote:
PAUSE <millis> is a OE 11.3 feature - if we implement it, we should activate it conditionally.

Greg Shah wrote:
I'm OK with this. We can handle the rounding vs preservation of the fractional part in the server and always pass millis value down to the client. I think we need to go ahead and implement this.

Constantin wrote:
I plan to use a legacy-system/wait-millis=true flag (in the directory.xml) to determine if READKEY and PAUSE will use the milliseconds. This rounding is done on server-side and client-side will be called with millis.

I prefer to check the Progress version number and automatically implement this if 11.3 or later. I don't want to add extra configuration where we can make a good default instead.

Greg Shah wrote:

I prefer to check the Progress version number and automatically implement this if 11.3 or later. I don't want to add extra configuration where we can make a good default instead.

You mean from p2j.cfg.xml? I don't recall having this saved in the directory. And p2j.cfg.xml is usually copied into the application jar only if runtime conversion is in use.

Also, my idea is for the ClientExports APIs to receive a millisecond value instead of seconds (as a double).
The server side will either round the received seconds or not, depending on that flag.

We need to adapt also the status message (for i.e. PAUSE) to reflect the millis, if they exist (3 decimals max).

CHOOSE ... PAUSE n also uses fractional seconds as millis.

You mean from p2j.cfg.xml?

No, this is runtime only. I was thinking of EnvironmentOps.getVersion(), which can read a version string from the directory but defaults to "11.6" at this time.

Greg Shah wrote:

You mean from p2j.cfg.xml?

No, this is runtime only. I was thinking of EnvironmentOps.getVersion(), which can read a version string from the directory but defaults to "11.6" at this time.

OK, I'll use that and configure other projects with this setting.

It is probably none of my business but I wonder why it is so complicated. PAUSE 1 is 1000 milliseconds, PAUSE 1.0 is also 1000 milliseconds and PAUSE 0.1 is 100 milliseconds. So I would think there are two overloads: one with an integer parameter and one with a double parameter, they both simply multiply by 1000 to get the number of milliseconds. I find it peculiar that the fractional part performs on the server and the millis on the client, why not just in one place? But like I said, it is none of my business really

Jurjen Dijkstra wrote:

I find it peculiar that the fractional part performs on the server and the millis on the client, why not just in one place?

The client is not receiving just the fractional part, it will receive the seconds converted to milliseconds (i.e. for a 1.9 seconds on server-side, client-side will receive 1900 millis).