7.3 Collecting the Data
The first task is deciding exactly what data we'll be collecting. Since we're using the humidity sensing circuit we developed in Section 4.4.3, we should briefly review its capabilities. The sensor used a 1-Wire chip as a digital front end to the physical humidity sensor. The humidity sensor's only output is an analog voltage. The 1-Wire chip provided analog to digital conversion as well as temperature readings. The HumiditySensor class we created to expose the sensor's functionality provides the following public methods.
public double getSensorRH() throws OneWireException public double getTrueRH() throws OneWireException public double getTemperature() throws OneWireException
Of the three readings we can obtain from the above methods, only two are likely to be interesting to a client: the temperature and the true relative humidity. The sensor relative humidity might be interesting for calibration purposes, but we'll ignore it here. We'll also want to put a time stamp on each reading so that clients can build logs and chart environmental change over time.
The next thing we need to decide is how to store the data samples. One obvious approach would be to write the data to a file. Each new entry could be appended to the end of the file. The advantage of using a file is that even if the system loses power, the log data is not lost. When power is restored and the application restarts, it can simply continue logging data samples by appending each sample to the end of the same file. If the system were down long enough to miss one or more samples, any client that downloads the file would be able to detect this by examining the time stamps. The downside to logging the data to a file is that DataLogger is running in a memory constrained environment. The file system, Java objects, and all system data structures live in the same memory space. If the log file grows too large, the application will likely terminate with an OutOfMemoryError or some other fatal exception. Special tricks would be required to ensure that the file didn't grow beyond a certain size. This is further complicated by the fact that we can't just truncate the file at a certain size by writing the latest sample over the sample at the end of the file. If a sample must be lost, it should be the oldest sample. In this sense, we really want something like a circular buffer. This can still be implemented with the file system using a RandomAccessFile, but it is too cumbersome for our example. For our purposes, it will be much simpler and more efficient to store the data in a Vector. Of course, if we just continue to add elements to the vector, we'll still run out of memory. But by using a Vector we can easily avoid this problem by removing the oldest sample, which will always be at index 0, before adding the new sample after the maximum sample count has been reached.
Note that if we assumed a constant connection to a network, we could structure DataLogger so that it just wrote all samples to a socket as they were collected. But we're building this application with the idea that the network isn't always available. This allows the logger to do all of its work without the network. Then when a client is interested in synching up with the logger, it can establish a connection with the server and collect the necessary data. The more "remote" the system is, the more important this ability becomes.
To keep the logging classes reasonably general purpose and reusable, we'll create an abstract class named LoggingDaemon to drive the data collection process. Ideally we don't want LoggingDaemon to have to be aware of what kind of data is being logged or the details of how it is acquired. To accomplish this isolation, LoggingDaemon defines the following abstract methods.
protected abstract Object captureSample(); protected abstract void writeLogEntry(Object sample, DataOutputStream dout) throws IOException;
Subclasses of LoggingDaemon implement the captureSample method to handle the details of collecting a single data sample. This sample must be encapsulated within an object because it will be stored in a Vector. The writeLogEntry method is used to write the individual fields contained in the sample object to the supplied instance of DataOutputStream.
LoggingDaemon's constructor is shown in Listing 7.4. The constructor requires the maximum number of samples to be held in the samples Vector along with the delay between consecutive samples. The maxSamples field is used to set the initial size of the Vector. The delay is input to the constructor as a number of seconds. The delay is converted to milliseconds so that it can be input directly into Thread's sleep method. Finally the LoggingDaemon thread is set to a daemon thread. This means that when the last non-daemon thread exits, LoggingDaemon will exit, along with any other daemon threads, allowing the process to terminate. We do this because there isn't any point in continuing to log data if there isn't a server running to allow clients to download it.
Listing 7.4 LoggingDaemon's constructor
import java.io.*;
import java.util.*;
public abstract class LoggingDaemon extends Thread {
private int maxSamples;
private int delay;
private Vector samples;
...
public LoggingDaemon(int maxSamples, int delay)
throws LoggingException {
this.maxSamples = maxSamples;
// Convert delay from seconds to milliseconds
this.delay = delay * 1000;
samples = new Vector(maxSamples);
this.setDaemon(true);
}
......
public void stopLogging() {
logEm = false;
}
}
LoggingDaemon's run method is shown in Listing 7.5. As long as the stopLogging method is not invoked, the run method spins in an infinite loop collecting data samples at the specified interval.
Listing 7.5 LoggingDaemon's run method
...
public void run() {
while (logEm) {
Object smp = captureSample();
if (smp != null) {
synchronized (samples) {
if (samples.size() == maxSamples) {
// Remove the oldest entry
samples.removeElementAt(0);
}
samples.addElement(smp);
}
}
try {
Thread.sleep(delay);
} catch (InterruptedException ie) {}
}
}
If the captureSample method returns null, there is no change in samples. The run method simply goes to sleep until it is time to try another sample. This is a rather simplistic mechanism for handling errors that occur during data collection, but it is appropriate for our application. Since every sample carries with it a time stamp, a client can determine that one or more samples were missed by simple analysis of the time stamps.
LoggingDaemon's writeLog method is shown in Listing 7.6. The writeLog method is invoked by the server when a client establishes a connection with the server, requesting a log of the recent data samples. The writeLog method simply enumerates samples, invoking writeLogEntry for every data sample contained within the Vector. The details of extracting and writing the actual field data contained within the sample object are left to the subclass.
Listing 7.6 LoggingDaemon's writeLog method
public void writeLog(DataOutputStream dout) throws IOException {
Vector sc = (Vector) samples.clone();
dout.writeInt(sc.size());
for (Enumeration e = sc.elements(); e.hasMoreElements(); ) {
writeLogEntry(e.nextElement(), dout);
}
}
Since we need to encapsulate the individual data readings within an object, we'll create a class named HumiditySample (shown in Listing 7.7). HumiditySample is just a thin wrapper on the sample data that provides public "get" methods for the individual fields. HumiditySample's constructor takes the readings attained using the HumiditySensor class and stores them in the temperature and relHumidity fields. It also time stamps the readings using the System.currentTimeMillis method, which returns the number of milliseconds between the current time and midnight, January 1, 1970. This is much simpler and faster for our purposes than storing the time stamp as a Date object. We can put the burden of converting the timeStamp value to humanly readable date and time on the client program. In the case that the client is written in Java, this job is trivial. It can simply pass the timeStamp value received to the Date constructor that takes the long value returned from currentTimeMillis. We'll make use of this in the next section, which presents a small sample client application.
Listing 7.7 HumiditySample
public class HumiditySample {
private double temperature;
private double relHumidity;
private long timeStamp;
public HumiditySample(double relHumidity, double temperature) {
this.temperature = temperature;
this.relHumidity = relHumidity;
timeStamp = System.currentTimeMillis();
}
public long getTimeStamp() {
return timeStamp;
}
public double getRelativeHumidity() {
return relHumidity;
}
public double getTemperature() {
return temperature;
}
}
Now that we have a simple framework for collecting, maintaining, and outputting a group of samples, we can create the class that performs the actual work of collecting individual samples. The class HumidityLogger, shown in Listing 7.8, extends LoggingDaemon and provides implementations for the captureSample and writeLogEntry methods.
Listing 7.8 HumidityLogger
import java.io.IOException;
import java.io.DataOutputStream;
import com.dalsemi.onewire.OneWireAccessProvider;
import com.dalsemi.onewire.adapter.DSPortAdapter;
import com.dalsemi.onewire.OneWireException;
public class HumidityLogger extends LoggingDaemon {
private HumiditySensor sensor;
private DSPortAdapter adapter;
public HumidityLogger(int maxSamples, int delay)
throws LoggingException {
super(maxSamples, delay);
try {
adapter = OneWireAccessProvider.getDefaultAdapter();
sensor = new HumiditySensor(adapter);
} catch (OneWireException owe) {
throw new LoggingException(
"Error creating Environmental Sensor:" +
owe.getMessage());
}
}
public Object captureSample() {
try {
adapter.beginExclusive(true);
double temp = sensor.getTemperature();
double humidity = sensor.getTrueRH();
return new HumiditySample(humidity, temp);
} catch (OneWireException owe) {
System.out.println("Error reading sensor");
owe.printStackTrace();
// No need to terminate app because of a failed reading
return null;
} finally {
adapter.endExclusive();
}
}
public void writeLogEntry(Object sample, DataOutputStream dout)
throws IOException {
dout.writeLong(((HumiditySample)sample).getTimeStamp());
dout.writeDouble(((HumiditySample)sample).getRelativeHumidity());
dout.writeDouble(((HumiditySample)sample).getTemperature());
}
}
HumidityLogger creates a new instance of the class HumiditySensor, which is used to perform the humidity and temperature measurements. When the captureSample method is invoked, it simply creates a new HumiditySample object to encapsulate the humidity and temperature values and returns that object to the caller, in this case the run method of LoggingDaemon. It is important that a transient error that could cause a failure while performing an individual measurement not cause the logging thread to terminate. If an exception is thrown while performing the measurements, it is caught, and null is returned.
Since the LoggingDaemon class doesn't know anything about the internal details of the data sample object—HumiditySample, in this case—it invokes writeLogEntry passing it a reference to the sample object and a DataOutputStream used to write the sample object's field information to the underlying socket. The writeLogEntry method extracts the time stamp, humidity, and temperature readings and writes them to the stream using the appropriate methods, writeLong and writeDouble, preserving their primitive types. It is assumed that the client will be using a DataInputStream for easy interpretation of the data.