jplace file

Placements in the jplace format (version 3), as written by pplacer, e.g. GTDB-Tk’s pplacer.class_level.bac120.json.

• tree: the reference tree with {edge} numbers, byte-identical to pplacer’s (labels quoted only when needed, branch lengths %g).

• placements: one record per distinct query sequence, {"p": [[...], ...], "nm": [[name, 1], ...]}, placements sorted by like-weight ratio. As in pplacer, queries whose pre-masked sequences are identical are placed once and share a record, names in input order.

• fields: distal_length, edge_num, like_weight_ratio, likelihood, pendant_length; with a reference package that has a taxonomy and seq_info, classification comes first: the tax_id of the placement’s edge (a string, or null), as pplacer writes it.

Numbers are printed as pplacer prints them (%.12g), so with gPlace’s bit-identical results every number is textually identical to pplacer’s. Whitespace, key order and the order of the query records are JSON-insignificant and may differ. As in pplacer, a non-finite value would be written as nan/inf.

Produced by

• place

• place_batch

• test

Example

{"tree": "(GB_GCA_016183025.1:0.31573{0},GB_GCA_036381125.1:0.34958{1},...)d__Bacteria{13713};", "placements":
  [
  {"p": [[0.0758007809161, 11051, 1, -13036489.6202, 0.0629997563924]], "nm": [["SPIREOTU_01109497", 1]]},
  ...
  ], "metadata": {"invocation": "gplace place ..."}, "version": 3, "fields":
  ["distal_length", "edge_num", "like_weight_ratio", "likelihood", "pendant_length"]
}