The adp parsing part

To access the functions provided, a reader must first be created. When creating the reader, it is possible to specify where the adb application is located.

reader := CreateReader("C:/Users/Test/AppData/Local/Android/Sdk/platform-tools/adb")

If adb is accessible via the PATH variable using adb, the path specification adbpath can be left empty:

reader := CreateReader("")

The 'Reader' interface now provides the following functions:

1. Read and parse memory info by using adb shell dumpsys meminfo
2. Read and parse running processes by using adb shell ps
3. Get rough trend information

Get memory information

This is the core functionality of the application. It first determines the process ID of the provided packagename with the help of the adb shell pidof call. Using the process ID and the adb shell dumpsys meminfo command, the memory information is retrieved and parsed. For common information groups, the essential information is converted to a key-value pair.

Scan(packagename string) (map[string]int, string, error)

Get running processes

The endpoint uses the adb shell ps -A -o NAME command to determine all active processes. It ignores process information enclosed by [ and ].

Packages() []string

Get rough trend

For the evaluation of the storage behavior, an initial indication of the trend of storage usage is required. For performance reasons, no complete history of all measured values is kept. Instead, a low-pass filter is used to determine trends.

Trend(key string, value int) (float64, float64, float64, float64)

The method is given a key and a current value. For the provided key it calculates four trends:

1. result #1: The change of the low-pass filter affected by the new value as follows: filtervalue = filtervalue * 0.999 + value * 0.001 (0.1%)
2. result #2: The change of the low-pass filter affected by the new value as follows: filtervalue = filtervalue * 0.99 + value * 0.01 (1%)
3. result #3: The change of the low-pass filter affected by the new value as follows: filtervalue = filtervalue * 0.9 + value * 0.1 (10%)
4. result #4: The change of the value, compared to the last measured one

This is not an exact trend calculation, especially since the trend information is lower the "slower" a low-pass filter reacts. Nevertheless, rough trends allow an overview of the memory behavior.