These objects read and write the file formats of the wider TDA ecosystem:
| object | formats |
|---|---|
CSV |
point clouds, full and lower-triangular distance matrices, persistence diagrams |
Ripser |
Ripser's point cloud, lower, upper, full and binary distance-matrix formats |
Dipha |
DIPHA's distance-matrix, image and persistence-diagram formats |
Gudhi |
GUDHI's OFF point clouds and persistence diagrams |
Perseus |
Perseus's cubical toplex and persistence-interval formats |
Each has raw readers (readPointCloud, readFullDistanceMatrix, ...) and readers that return something to compute with (readEuclideanMetricSpace, readExplicitMetricSpace, readCubicalGridStream, ...):
import scala.language.experimental.modularity
import org.appliedtopology.tda4j.*
val points = CSV.readPointCloud("points.csv")
val diagram = Persistence(points)
val distances = Ripser.readLowerDistanceExplicitMetricSpace("distances.txt")
val fromDistances = Persistence(distances)
val image = Persistence(Perseus.readCubicalToplex("image.txt"))
CSV.writePersistenceDiagram("diagram.csv", diagram.bars)
Two details if you compare with another tool: Ripser's binary distance matrices are 32-bit floats, and DIPHA's and Perseus's grids list the first axis fastest (the readers and writers convert for you). A Perseus value of -1 is a missing cell, read as NaN: masked, not in the complex, sublevel or superlevel. Writing a grid writes NaN as -1 and refuses an infinite value (the format has none).
A whole result of the MATLAB interface or the command line, with its representatives and the points or image it was computed from, is saved as one JSON file by result.save(path) (--save-result) and read back by TDA4j.loadResult(path): see saved results.