tsubasa

Overview

This package is used for climate network construction. More details can be found from this paper.

Install

Download this package from the GitHub repository and then import it locally.

Prerequisite: NetCDF C Library must be installed firstly.

Documentation

Initialization

The init functions are in init.go.

1. Init(void) void

   Initiate all required global variables and data structures. All data will be stored in and loaded from the memory. It must be called before any other operations.

2. InitDB(username string, password string) void

   Initiate all required global variables and data structures for disk storage. Data will be stored onto disk. The statistics computed from the originial data could be retrieved and reused from PostgreSQL database on the machine. It also must be called before any other operations.

3. InitMatrix(void) void

   Initiate or reset all values of matrices to zero.

Data Loading

It can load data directly from NetCDF files (ended with .nc). The corresponding functions are in readfiles.go.

1. ReadFile(fileName string) void

   Read the given NetCDF file.

2. ReadFiles(directoryName string) error

   Read all NetCDF files from the specific directory.

3. ReadFileByLocation(fileName string, locationRangeFile string) void

   Read the given NetCDF file. The second argument should be a text file to limit the geographical ranges, which contains one line with 4 integers separated by ",". For instance, 100,150,-10,10 denotes minimum longitude, maximum longitude, minimum latitude, and maximum latitude, respectively.

4. ReadFilesByLocation(directoryName string, locationRangeFile string) error

   Read all NetCDF files from the specific directory. The second argument should be a text file to limit the geographical ranges, which contains one line with 4 integers separated by ",". For instance, 100,150,-10,10 denotes minimum longitude, maximum longitude, minimum latitude, and maximum latitude, respectively.

Information Acquisition

These functions are from utils.go and tsubasago.go.

1. GetTimeSeriesNum(void) int

   Get the total number of time series

2. GetTimeSeriesLength(void) int

   Get the length of time series. Assume all time series has the same length.

3. GetBasicWindowSize(void) int

   Get the size of basic window.

4. GetMatrix(void) []int

   Get the weighted correlation matrix. The return value is an integer array, which transferred from the N * N matrix to a 1 * (N^2) vector.

5. GetRealMatrix(void) []float64

   Get the unweighted correlation matrix. The return value is a float array, which transferred from the N * N matrix to a 1 * (N^2) vector.

6. GetDataMapInfo(void) int

   Check if each time series has the same length. Meanwhile, return the length of time series.

Parameters Setting

These functions are from utils.go.

1. SetBasicWindowSize(size int) void

   Set the size of basic window.

Computation

These functions are used to compute the intermediate statistics and the final correlation matrices. They are from tsubasago.go.

Note: The package does not provide exposed methods for DFT methods.

1. DirectCompute(thres float64, start int, end int) []int

   Direct in-memory computation with parallel computing. The number of Goroutines depends on the number of CPU obtained from runtime.NumCPU(). thred usually is set to the value between 0.6 and 0.95. start and end defines the length of time series.

2. Sketch(void) string

   In-memory sketch with parallel computing. The number of Goroutines depends on the number of CPU obtained from runtime.NumCPU(). It returns the total time represented by a string with a time unit.

3. Query(thres float64, queryStart int, queryEnd int) []int

   In-memory Query with parallel computing. The number of Goroutines depends on the number of CPU obtained from runtime.NumCPU(). queryStart and queryEnd are the start and end index of basic window, which makes up to the query window. It returns the weighted correlation matrix. It is an integer array, which transferred from the N * N matrix to a 1 * (N^2) vector. (N is the number of time series)

4. SketchInDB(writersNum int) void

   In-DB sketch with parallel computing. The number of Goroutines depends on the number of CPU obtained from runtime.NumCPU(). The writersNum should be at least 1 and smaller than the half of the total Goroutines (NCPU).

5. QueryInDB(thres float64, start int, end int, granularity int, writersNum int) []int

   Query with parallel computing. It reads the statistics from PostgreSQL database. The number of Goroutines depends on the number of CPU obtained from runtime.NumCPU(). granularity is the basic window size. writersNum should be the same as the value in SketchInDB(writersNum int) void. It returns the weighted correlation matrix. It is an integer array, which transferred from the N * N matrix to a 1 * (N^2) vector. (N is the number of time series)

6. ResetSketch(writersNum int) void

   Reset to the initial state, which means cleanning all results after calling SketchInDB(writersNum int) void.

Example

A simple use case is provided below,

func main() {
	// Initialization
	tsubasa.Init()

	// Read data from a NetCDF file
	tsubasa.ReadFileByLocation("../data.nc", "range.txt")

	// Get time series length
	length := tsubasa.GetTimeSeriesLength()

	// Set basic window size
	tsubasa.SetBasicWindowSize(30)

	tsubasa.Sketch()
	tsubada.Query(0.75, 0, int(length/30) - 1)
}