Split DiskIOMeter into disk IO time and rate sub-meters

[Explorer09]

Split the DiskIOMeter into two, DiskIOTimeMeter and DiskIORateMeter, showing the time and the rate separately.

The name DiskIOMeter is retained but now shows the combined view of time and rate, similar to MemorySwapMeter.

This PR depends on #1721 (merged), which introduces an isPercentChart property to meters. Note this isPercentChart property is different between sub-meters, and it could explain the reason why the split is worth it.

[BenBE]

What's currently still open for this PR to become ready?

[Explorer09]

What's currently still open for this PR to become ready?

I originally made this PR as a "proof of concept" and to grab early comments about the meters' look and functions.

• I also intended that DiskIOTimeMeter will include the number of disks in the display, so that the greater-than-100% disk time can be understood without reading more context or manual, however I don't have an idea on how to implement that well. I think I can postpone that part into a separate PR. I just leave a reminder here. (Update: This one is done in the new revision.)

[BenBE]

Just did a quick check with the design and I think the combined meter should show IO rates left, IO Time used on the right. Else you get quite a big gap with just the percentage shown in text mode.

[Explorer09]

Just did a quick check with the design and I think the combined meter should show IO rates left, IO Time used on the right. Else you get quite a big gap with just the percentage shown in text mode.

Makes sense. Perhaps another good reason is that people might look at the I/O rate graph more often than the time/utilisation percentage. (Besides, the I/O rate graph can support two channels while the percentage graph cannot.)

[Explorer09]

@BenBE On the topic of avoiding double assignment in code style, I found there are three other files that had this "double assignment" syntax when initializing data. Not sure if you want these to be changed as well, but I'm reluctant to file a new PR for this minor style change:

--- a/CPUMeter.c
+++ b/CPUMeter.c
@@ -241,9 +241,10 @@ static void CPUMeterCommonInit(Meter* this) {
 
    CPUMeterData* data = this->meterData;
    if (!data) {
-      data = this->meterData = xMalloc(sizeof(CPUMeterData));
+      data = xMalloc(sizeof(CPUMeterData));
       data->cpus = this->host->existingCPUs;
       data->meters = count ? xCalloc(count, sizeof(Meter*)) : NULL;
+      this->meterData = data;
    }
 
    Meter** meters = data->meters;
--- a/Machine.c
+++ b/Machine.c
@@ -79,9 +79,11 @@ void Machine_populateTablesFromSettings(Machine* this, Settings* settings, Table
 
    for (size_t i = 0; i < settings->nScreens; i++) {
       ScreenSettings* ss = settings->screens[i];
+
+      if (!ss->table)
+         ss->table = processTable;
+
       Table* table = ss->table;
-      if (!table)
-         table = ss->table = processTable;
       if (i == 0)
          this->activeTable = table;
 
--- a/MemorySwapMeter.c
+++ b/MemorySwapMeter.c
@@ -47,11 +47,10 @@ static void MemorySwapMeter_draw(Meter* this, int x, int y, int w) {
 }
 
 static void MemorySwapMeter_init(Meter* this) {
-   MemorySwapMeterData* data = this->meterData;
+   if (!this->meterData)
+      this->meterData = xCalloc(1, sizeof(MemorySwapMeterData));
 
-   if (!data) {
-      data = this->meterData = xCalloc(1, sizeof(MemorySwapMeterData));
-   }
+   MemorySwapMeterData* data = this->meterData;
 
    if (!data->memoryMeter)
       data->memoryMeter = Meter_new(this->host, 0, (const MeterClass*) Class(MemoryMeter));

[BenBE]

@Explorer09 Patch applied on main with slight modifications.

[BenBE]

Where does the 40% come from?

[Explorer09]

e518459 I would let @cgzones answer this.