tpattern 0.1.9__tar.gz → 0.1.10__tar.gz

{tpattern-0.1.9/tpattern.egg-info → tpattern-0.1.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tpattern
-Version: 0.1.9
+Version: 0.1.10
 Summary: Open, auditable temporal & sequential pattern analysis (T-patterns) for sport and behaviour
 Author-email: Andrew Callaway <acallaway@bournemouth.ac.uk>
 License-Expression: MIT

{tpattern-0.1.9 → tpattern-0.1.10}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "tpattern"
-version = "0.1.9"
+version = "0.1.10"
 description = "Open, auditable temporal & sequential pattern analysis (T-patterns) for sport and behaviour"
 readme = "README.md"
 requires-python = ">=3.9"

tpattern-0.1.10/tests/test_read_table_validation.py

@@ -0,0 +1,121 @@
+"""Data-check tests for read_table: the E/W/I codes, and that the clean path
+stays silent and unchanged. See tpattern/io.py CHECKS."""
+import pytest
+from tpattern import read_table, DataFormatError, CHECKS, check_reference
+
+
+def write(tmp_path, name, text):
+    p = tmp_path / name
+    p.write_text(text)
+    return str(p)
+
+
+# ---- E codes: raise, nothing loaded ----
+
+def test_E01_missing_column(tmp_path):
+    f = write(tmp_path, "m.csv", "observation,event,time\np1,pass,0\np1,shot,3\n")
+    with pytest.raises(DataFormatError) as e:
+        read_table(f)
+    assert "E01" in str(e.value) and "start" in str(e.value)
+
+
+def test_E01_suggests_close_column(tmp_path):
+    f = write(tmp_path, "m.csv", "observation,event,Start\np1,pass,0\np1,shot,3\n")
+    with pytest.raises(DataFormatError) as e:
+        read_table(f)
+    assert "Start" in str(e.value)  # suggested the near-match
+
+
+def test_E02_wrong_separator(tmp_path):
+    f = write(tmp_path, "s.csv", "observation;event;start\np1;pass;0\np1;shot;3\n")
+    with pytest.raises(DataFormatError) as e:
+        read_table(f)
+    assert "E02" in str(e.value)
+
+
+def test_E03_header_only(tmp_path):
+    f = write(tmp_path, "h.csv", "observation,event,start\n")
+    with pytest.raises(DataFormatError) as e:
+        read_table(f)
+    assert "E03" in str(e.value)
+
+
+def test_E04_all_rows_blank(tmp_path):
+    f = write(tmp_path, "b.csv", "observation,event,start\n,pass,0\n,shot,3\n")
+    with pytest.raises(DataFormatError) as e:
+        read_table(f)
+    assert "E04" in str(e.value)
+
+
+def test_E05_no_pattern_capable_observation(tmp_path):
+    # every observation has exactly one event
+    f = write(tmp_path, "one.csv",
+              "observation,event,start\np1,pass,0\np2,shot,0\np3,cross,0\n")
+    with pytest.raises(DataFormatError) as e:
+        read_table(f)
+    assert "E05" in str(e.value)
+
+
+# ---- W codes: load anyway, reported ----
+
+def test_W10_non_numeric_times_dropped(tmp_path, capsys):
+    f = write(tmp_path, "c.csv",
+              "observation,event,start\np1,pass,0\np1,cross,01:03\np1,shot,5\n"
+              "p2,pass,1\np2,shot,4\n")
+    obs = read_table(f)                       # default report: a W fired, so it prints
+    out = capsys.readouterr().out
+    assert "W10" in out and "left out" in out
+    # the one good p1 row + p2 both loaded; the mm:ss row dropped
+    assert sum(len(o.events) for o in obs) == 4
+
+
+def test_W13_near_duplicate_labels(tmp_path, capsys):
+    f = write(tmp_path, "d.csv",
+              "observation,event,start\np1,Pass,0\np1,pass,1\np1,shot,2\n"
+              "p2,Pass,0\np2,shot,2\n")
+    read_table(f)
+    assert "W13" in capsys.readouterr().out
+
+
+def test_clean_file_is_silent(tmp_path, capsys):
+    f = write(tmp_path, "ok.csv",
+              "observation,event,start\np1,pass,0\np1,shot,2\np2,pass,0\np2,shot,3\n")
+    obs = read_table(f)
+    assert capsys.readouterr().out == ""       # nothing to warn about
+    assert len(obs) == 2
+
+
+# ---- I codes: only shown with report="all" ----
+
+def test_I20_effective_sample_only_with_all(tmp_path, capsys):
+    f = write(tmp_path, "e.csv",
+              "observation,event,start\np1,pass,0\np1,shot,2\np2,pass,0\n")  # p2 has 1 event
+    read_table(f)                              # default: I-only, stays silent
+    assert capsys.readouterr().out == ""
+    read_table(f, report="all")                # report='all' surfaces the info note
+    assert "I20" in capsys.readouterr().out
+
+
+def test_report_false_silences_warnings(tmp_path, capsys):
+    f = write(tmp_path, "c.csv",
+              "observation,event,start\np1,pass,0\np1,x,bad\np1,shot,5\np2,a,0\np2,b,1\n")
+    read_table(f, report=False)
+    assert capsys.readouterr().out == ""
+
+
+# ---- BOM (Excel) handled ----
+
+def test_utf8_bom_is_stripped(tmp_path):
+    p = tmp_path / "bom.csv"
+    p.write_bytes("observation,event,start\np1,pass,0\np1,shot,2\n".encode("utf-8"))
+    obs = read_table(str(p))                   # must not see the BOM as part of the column name
+    assert len(obs) == 1 and len(obs[0].events) == 2
+
+
+# ---- registry integrity (Appendix A generates from this) ----
+
+def test_check_reference_covers_every_code():
+    rows = check_reference()
+    assert {r["code"] for r in rows} == set(CHECKS)
+    for r in rows:
+        assert r["condition"] and r["fix"] and r["severity"]

{tpattern-0.1.9 → tpattern-0.1.10}/tpattern/__init__.py

@@ -6,9 +6,10 @@ behavior: T-patterns and their detection. Behavior Research Methods,
 Instruments, & Computers, 32(1), 93-110.
 """
 
-__version__ = "0.1.9"   # defined before submodule imports (methods.py reads it)
+__version__ = "0.1.10"   # defined before submodule imports (methods.py reads it)
 
-from .io import Observation, read_observation, read_sample, read_table, read_sofcoder
+from .io import (Observation, read_observation, read_sample, read_table,
+                 read_sofcoder, DataFormatError, check_reference, CHECKS)
 from .pattern import Instance, Pattern
 from .ci import find_critical_interval, CIResult
 from .detect import Config, Engine
@@ -24,6 +25,7 @@ from .contrast import group_contrast, contrast_items
 
 __all__ = [
     "Observation", "read_observation", "read_sample", "read_table", "read_sofcoder",
+    "DataFormatError", "check_reference", "CHECKS",
     "Instance", "Pattern",
     "find_critical_interval", "CIResult",
     "Config", "Engine",

tpattern-0.1.10/tpattern/io.py

@@ -0,0 +1,679 @@
+"""
+Reading your event data in.
+===========================
+
+Almost everyone wants one function: :func:`read_table`. Give it a plain table with
+one row per event, and it returns the list of *observations* the rest of the
+library analyses::
+
+    from tpattern import read_table
+    observations = read_table("my_events.csv")
+
+Your table needs three columns — an **observation** id (which session/rally/bout/
+possession the event belongs to), an **event** label (what happened), and a
+**start** time (when). Anything else is optional (see SCHEMA.md).
+
+An *observation* is one self-contained time-line: a single possession, rally, bout
+or match. Patterns are counted across all of your observations, but one occurrence
+of a pattern can never span two of them — so getting the observation column right
+matters. Several events may share the same time; the engine treats same-time events
+as an unordered set at that instant (use ``Config.min_lag`` to require a real gap
+instead).
+
+**Reading your own, un-cleaned data.** Real exports are messy. Rather than fail on
+the first bad row or, worse, drop rows silently, :func:`read_table` runs a set of
+checks and tells you — in plain language — what it found, how much of your file was
+affected, and what to do about it. Each check has a short code (``E01``, ``W05`` …)
+explained in the tutorial's Appendix A and in :data:`CHECKS`. A code beginning
+``E`` stops the read (something must be fixed first); ``W`` and ``I`` never stop it
+— the data loads and the notes are there if you want to tidy up.
+
+:func:`read_sample` / :func:`read_observation` are an older reader for THEME's
+tab-separated files (one file per observation, ``time<TAB>event`` rows, with ``:``
+and ``&`` marking the window). :func:`read_sofcoder` reads SOF-CODER/THEME
+``DATANAME, T, Events`` files.
+"""
+
+from __future__ import annotations
+
+import csv
+import difflib
+from collections import defaultdict, Counter
+from dataclasses import dataclass, field
+from pathlib import Path
+
+
+# The two reserved symbols that mark the observation window rather than a real
+# event, in the older THEME/SOF-CODER file formats.
+START_MARKER = ":"
+END_MARKER = "&"
+
+
+@dataclass
+class Observation:
+    """One time-line: all the events of a single observation.
+
+    Attributes
+    ----------
+    name : str
+        Which observation this is (its id), so a pattern can be traced back to a
+        specific session/rally/bout.
+    start, end : int
+        The observation window [start, end] in milliseconds. ``T = end - start`` is
+        the length of the window, used as the denominator of the NX/T baseline (how
+        likely an event is to fall in a given slice of time by chance).
+    events : list[tuple[int, str]]
+        (time, event label) pairs, sorted by time, with any window markers removed.
+    """
+
+    name: str
+    start: int
+    end: int
+    events: list[tuple[int, str]] = field(default_factory=list)
+
+    @property
+    def T(self) -> int:
+        """Length of the observation window, in time units (ms)."""
+        return self.end - self.start
+
+
+# ---------------------------------------------------------------------------
+# Data checks: the codes, and the human-readable report.
+# ---------------------------------------------------------------------------
+
+class DataFormatError(ValueError):
+    """Raised when a file cannot be read as it stands (an ``E`` check).
+
+    The message *is* the plain-language report: what is wrong, and how to fix it.
+    ``W`` and ``I`` checks never raise — they are reported and the data still loads.
+    """
+
+
+# The single source of truth for every check. The runtime report and the tutorial's
+# Appendix A are both built from this, so they can never drift apart. `severity`:
+# 'E' stops the read (must fix first), 'W' is a warning (loads anyway, worth tidying),
+# 'I' is information (loads anyway, nothing to fix). `condition` says when it fires;
+# `fix` says what to do; `example` shows it concretely (a code snippet, or a data
+# before -> after). Runtime adds the specifics (counts, percentages, example rows).
+CHECKS: dict[str, dict[str, str]] = {
+    "E01": {"severity": "E", "condition": "A column the analysis needs is not in the file.",
+            "fix": "Rename the column, or tell read_table its name.",
+            "example": "read_table(\"events.csv\", start=\"Start Time\")"},
+    "E02": {"severity": "E", "condition": "The file read as a single column, so it is probably not comma-separated.",
+            "fix": "Re-save it as a comma-separated CSV, or pass the separator.",
+            "example": "read_table(\"events.csv\", sep=\";\")"},
+    "E03": {"severity": "E", "condition": "The file has a header row but no data under it (or is empty).",
+            "fix": "Check the export actually contains events, then re-run.",
+            "example": "a data row looks like: p1,pass,0"},
+    "E04": {"severity": "E", "condition": "Every row was unusable, so there was nothing to load.",
+            "fix": "Open the file and check the observation, event and start columns are filled in.",
+            "example": "p1,pass,0 (a value in observation, event and start)"},
+    "E05": {"severity": "E", "condition": "No observation has at least two events, so no pattern can form.",
+            "fix": "Check the observation column groups events correctly; a pattern needs two or more events in the same observation.",
+            "example": "two rows sharing an id: p1,pass,0 and p1,shot,3"},
+    "W10": {"severity": "W", "condition": "Some start times are not numbers, so those rows were left out.",
+            "fix": "Times must be numbers. If yours is a mm:ss clock, convert it to seconds first.",
+            "example": "01:23 -> 83"},
+    "W11": {"severity": "W", "condition": "Some rows have a blank observation id and were left out.",
+            "fix": "Fill in the observation column, or remove the empty rows.",
+            "example": ",pass,0 -> p1,pass,0"},
+    "W12": {"severity": "W", "condition": "Some rows have a blank event label and were left out.",
+            "fix": "Fill in the event column, or remove the empty rows.",
+            "example": "p1,,0 -> p1,pass,0"},
+    "W13": {"severity": "W", "condition": "Two or more event labels differ only by capitalisation or spacing.",
+            "fix": "They are counted as different events. Merge them if they mean the same thing.",
+            "example": "Pass and pass -> both Pass"},
+    "W14": {"severity": "W", "condition": "Some events fall outside their observation's window, or a window has end at or before start.",
+            "fix": "Check the obs_start / obs_end columns; the window should contain all of the observation's events.",
+            "example": "event at 620 with window 0-600 -> set obs_end >= 620"},
+    "W15": {"severity": "W", "condition": "The time values look like they might be in the other unit.",
+            "fix": "Check time_unit: pass \"s\" for seconds or \"ms\" for milliseconds.",
+            "example": "read_table(\"events.csv\", time_unit=\"s\")"},
+    "I20": {"severity": "I", "condition": "Some observations have fewer than two events.",
+            "fix": "None needed. They set the baseline event rates but cannot themselves contain a pattern; the Methods text reports both counts.",
+            "example": "Methods reports \"163 observations, 131 with >=2 events\""},
+    "I21": {"severity": "I", "condition": "The file contains only one observation.",
+            "fix": "None needed, but check it is intended: patterns are counted across observations, so a single time-line will detect little.",
+            "example": "is the observation column accidentally a constant?"},
+    "I22": {"severity": "I", "condition": "One event type makes up a large share of all events.",
+            "fix": "None needed. A very common event can link to almost everything; the advisor may suggest setting it aside from pattern-building.",
+            "example": "exclude_events=[\"Interception\"]"},
+    "I23": {"severity": "I", "condition": "A high fraction of neighbouring events share an identical time.",
+            "fix": "None needed. The advisor will likely suggest requiring a genuine gap so co-timed events are read as co-occurrence, not sequence.",
+            "example": "Config(min_lag=1)"},
+    "I24": {"severity": "I", "condition": "Some rows are exact duplicates (same observation, event and time).",
+            "fix": "None needed by default: identical same-time events are counted once. If they are double entries and you turned that collapsing off, remove the copies.",
+            "example": "keep collapse_duplicates on (the default)"},
+}
+
+
+@dataclass
+class _Finding:
+    """One thing a check turned up, with the specifics filled in at runtime."""
+    code: str
+    detail: str            # the specifics: counts, percentages, an example
+
+    @property
+    def severity(self) -> str:
+        return CHECKS[self.code]["severity"]
+
+    @property
+    def fix(self) -> str:
+        return CHECKS[self.code]["fix"]
+
+    @property
+    def example(self) -> str:
+        return CHECKS[self.code].get("example", "")
+
+
+class CheckReport:
+    """Collects findings for one file and renders the plain-language report.
+
+    Access ``.findings`` for the raw list, ``.errors`` for the blocking ones, and
+    ``str(report)`` for the human-readable text. ``read_table`` attaches the report
+    to what it prints; a blocking finding is raised as :class:`DataFormatError`.
+    """
+
+    def __init__(self, filename: str, rows_read: int):
+        self.filename = filename
+        self.rows_read = rows_read          # data rows (header excluded)
+        self.rows_used = rows_read
+        self.n_observations = 0
+        self.findings: list[_Finding] = []
+
+    def add(self, code: str, detail: str) -> None:
+        self.findings.append(_Finding(code, detail))
+
+    @property
+    def errors(self) -> list[_Finding]:
+        return [f for f in self.findings if f.severity == "E"]
+
+    def _pct(self, n: int) -> str:
+        base = self.rows_read or 1
+        return f"{n} ({n / base:.1%} of {self.rows_read})"
+
+    def render(self) -> str:
+        lines: list[str] = []
+        if self.errors:
+            lines.append(f"Data check — {self.filename}   ✗ couldn't load the file")
+            lines.append("Nothing was loaded yet. Fix and re-run:")
+            for f in self.errors:
+                lines.append(f"  • {f.detail} ({f.code})")
+                eg = f"  e.g. {f.example}" if f.example else ""
+                lines.append(f"      → {f.fix}{eg}")
+            return "\n".join(lines)
+
+        # Loaded successfully; W/I notes are optional.
+        lines.append(f"Data check — {self.filename}   ✓ loaded")
+        if self.rows_used < self.rows_read:
+            left = self.rows_read - self.rows_used
+            lines.append(f"  Loaded {self.rows_used} of {self.rows_read} rows "
+                         f"({left} left out, {left / (self.rows_read or 1):.1%}) "
+                         f"across {self.n_observations} observations.")
+        else:
+            lines.append(f"  Loaded all {self.rows_used} rows across "
+                         f"{self.n_observations} observations.")
+        notes = [f for f in self.findings if f.severity in ("W", "I")]
+        if notes:
+            lines.append("  Your data is ready. The notes below are optional — "
+                         "nothing here stopped the load:")
+            for f in notes:
+                lines.append(f"    • {f.detail} ({f.code})")
+                if f.severity == "W":
+                    eg = f"  e.g. {f.example}" if f.example else ""
+                    lines.append(f"        → {f.fix}{eg}")
+        else:
+            lines.append("  Your data is ready. No issues found.")
+        return "\n".join(lines)
+
+    def __str__(self) -> str:      # so print(report) just works
+        return self.render()
+
+
+def _closest_column(wanted: str, available) -> str | None:
+    """Best guess at which real column a missing required one was meant to be."""
+    hit = difflib.get_close_matches(wanted, list(available), n=1, cutoff=0.6)
+    if hit:
+        return hit[0]
+    low = {c.lower(): c for c in available}
+    return low.get(wanted.lower())
+
+
+def _norm_label(code: str) -> str:
+    """Fold a label to compare look-alikes: lower-case, single-spaced, trimmed."""
+    return " ".join(code.lower().split())
+
+
+def _example_rows(rows: list[int]) -> str:
+    """Render a couple of example row numbers for a message."""
+    shown = ", ".join(f"row {r}" for r in rows[:2])
+    return shown
+
+
+def read_table(path: str | Path, *, observation: str = "observation",
+               event: str = "event", start: str = "start",
+               end: str | None = "end", obs_start: str | None = "obs_start",
+               obs_end: str | None = "obs_end", build_event_from: list[str] | None = None,
+               time_unit: str = "s", sep: str = ",",
+               check: bool = True, report: bool = True) -> list[Observation]:
+    """Read a plain event table (one row per event) into a list of observations.
+
+    This is the software-agnostic way in: any tool — OpenTag, SportsCode, a
+    spreadsheet — can export these columns, and the rest of the library is the same
+    whatever produced them. Rows are grouped by the ``observation`` column into one
+    :class:`Observation` each.
+
+    Parameters
+    ----------
+    observation, event, start : str
+        The names of the three required columns in *your* file. If your export uses
+        different names, pass them here (e.g. ``start="Start Time"``).
+    end : str, optional
+        A column giving each event's end time. If it is absent the observation
+        window is taken from the earliest and latest event.
+    obs_start, obs_end : str, optional
+        Columns giving the observation *window* (the same value for every row of an
+        observation). These default to the usual names, so a file that includes them
+        is read correctly with no extra arguments, and a file without them simply
+        falls back to the first and last event. Give a window when the observation
+        has real bounds beyond its first and last event — a possession or rally
+        usually does. The window length ``T = obs_end - obs_start`` is the
+        denominator of the NX/T baseline, so taking it from the events instead shifts
+        every baseline probability.
+    build_event_from : list[str], optional
+        Build the event label by joining these columns with '_', for data that keeps
+        the label split across descriptor columns, instead of reading ``event``.
+    time_unit : {'s', 'ms'}
+        The unit of your time columns. Times are held internally as whole
+        milliseconds, so seconds are multiplied by 1000.
+    check : bool
+        Run the data checks (default on). See :data:`CHECKS` and Appendix A.
+    report : bool
+        Print the plain-language report when a check finds something (default on).
+        A clean file prints nothing. Blocking (``E``) problems are always raised as
+        :class:`DataFormatError` whatever this is set to.
+
+    Raises
+    ------
+    DataFormatError
+        If the file cannot be read as it stands (any ``E`` check).
+    """
+    if time_unit not in ("s", "ms"):
+        raise ValueError(f"time_unit must be 's' or 'ms', got {time_unit!r}")
+    path = Path(path)
+    scale = 1000 if time_unit == "s" else 1
+
+    # utf-8-sig quietly strips the byte-order mark Excel puts on the first column
+    # name, which would otherwise make the observation column look "missing".
+    with open(path, newline="", encoding="utf-8-sig") as fh:
+        reader = csv.DictReader(fh, delimiter=sep)
+        fieldnames = list(reader.fieldnames or [])
+        rows = list(reader)
+
+    rep = CheckReport(path.name, rows_read=len(rows))
+    cols = set(fieldnames)
+
+    def finish_or_raise():
+        """Emit the report and, if anything blocks, raise it.
+
+        A blocking (E) finding is always raised. Otherwise the plain-language report
+        is printed only when there is something the user should act on — any warning
+        (W) — so a clean file, or one with information-only notes, stays quiet.
+        Pass ``report="all"`` to print the report whenever there is any note at all
+        (what the guided notebook does, to surface the information notes too).
+        """
+        if check and rep.errors:
+            raise DataFormatError("\n" + rep.render())
+        if not (check and report) or not rep.findings:
+            return
+        if report == "all" or any(f.severity == "W" for f in rep.findings):
+            print(rep.render())
+
+    # --- E02: the whole file read as one column => wrong separator ---
+    if check and len(fieldnames) == 1 and (";" in fieldnames[0] or "\t" in fieldnames[0]):
+        guess = ";" if ";" in fieldnames[0] else "a tab"
+        rep.add("E02", f"the file read as a single column ({fieldnames[0]!r}); it "
+                       f"looks separated by {guess}, not commas")
+        finish_or_raise()
+
+    # --- E01: required columns present? (with a best-guess suggestion) ---
+    code_cols = set(build_event_from) if build_event_from else {event}
+    required = {observation: observation, start: start}
+    required.update({c: c for c in code_cols})
+    missing = [c for c in required if c not in cols]
+    if check and missing:
+        for c in missing:
+            guess = _closest_column(c, cols)
+            hint = f"; your file has {guess!r} — did you mean that?" if guess else ""
+            rep.add("E01", f"no column called {c!r}. Columns found: "
+                           f"{sorted(cols)}{hint}")
+        finish_or_raise()
+
+    # --- E03: header but no data ---
+    if check and not rows:
+        rep.add("E03", "the file has a header row but no events under it")
+        finish_or_raise()
+
+    # --- parse rows, recording (not hiding) anything we cannot use ---
+    rows_by_obs: dict[str, list] = defaultdict(list)
+    ends_by_obs: dict[str, list] = defaultdict(list)
+    window_by_obs: dict[str, tuple] = {}
+    seen_rows: set = set()
+
+    blank_obs: list[int] = []
+    blank_event: list[int] = []
+    bad_time: list[int] = []
+    bad_time_examples: list[str] = []
+    dup_rows: list[int] = []
+    used = 0
+
+    for i, r in enumerate(rows, start=2):        # row 1 is the header, so data starts at 2
+        obs_id = (r.get(observation) or "").strip()
+        if not obs_id:
+            blank_obs.append(i)
+            continue
+        if build_event_from:
+            parts = [(r.get(c) or "").strip() for c in build_event_from]
+            code = "_".join(p for p in parts if p)
+        else:
+            code = (r.get(event) or "").strip()
+        if not code:
+            blank_event.append(i)
+            continue
+        try:
+            t = int(round(float(r[start]) * scale))
+        except (ValueError, TypeError, KeyError):
+            bad_time.append(i)
+            if len(bad_time_examples) < 2:
+                bad_time_examples.append(f"row {i} '{str(r.get(start))[:20]}'")
+            continue
+
+        key = (obs_id, code, t)
+        if key in seen_rows:
+            dup_rows.append(i)
+        seen_rows.add(key)
+
+        rows_by_obs[obs_id].append((t, code))
+        used += 1
+        if end and r.get(end) not in (None, ""):
+            try:
+                ends_by_obs[obs_id].append(int(round(float(r[end]) * scale)))
+            except ValueError:
+                pass
+        if obs_start and obs_end and obs_id not in window_by_obs:
+            try:
+                window_by_obs[obs_id] = (int(round(float(r[obs_start]) * scale)),
+                                         int(round(float(r[obs_end]) * scale)))
+            except (ValueError, TypeError, KeyError):
+                pass
+
+    rep.rows_used = used
+
+    # --- build the observations (identical to the clean-data path) ---
+    sample: list[Observation] = []
+    bad_window = 0
+    out_of_window = 0
+    for obs_id, evs in rows_by_obs.items():
+        evs.sort(key=lambda e: e[0])
+        if obs_id in window_by_obs:
+            o_start, o_end = window_by_obs[obs_id]
+        else:
+            o_start = evs[0][0]
+            o_end = max(ends_by_obs[obs_id]) if ends_by_obs.get(obs_id) else evs[-1][0]
+        if o_end < o_start:            # inverted window (a zero-length one is fine)
+            bad_window += 1
+        else:
+            out_of_window += sum(1 for t, _ in evs if t < o_start or t > o_end)
+        sample.append(Observation(name=obs_id, start=o_start, end=o_end, events=evs))
+    rep.n_observations = len(sample)
+
+    # --- checks that need the parsed data ------------------------------------
+    if check:
+        # E04 / E05: nothing usable, or nothing that can host a pattern
+        if not sample:
+            reasons = []
+            if blank_obs:
+                reasons.append(f"{len(blank_obs)} had a blank {observation}")
+            if blank_event:
+                reasons.append(f"{len(blank_event)} had a blank {event}")
+            if bad_time:
+                reasons.append(f"{len(bad_time)} had a start time that isn't a number")
+            why = "; ".join(reasons) if reasons else "no rows had all three of "\
+                  f"{observation}, {event} and {start}"
+            rep.add("E04", f"every row was left out ({why})")
+            finish_or_raise()
+        elif not any(len(o.events) >= 2 for o in sample):
+            rep.add("E05", "no observation contains two or more events, so no "
+                           "pattern can form")
+            finish_or_raise()
+
+        # W10-W12: rows left out
+        if bad_time:
+            eg = "; ".join(bad_time_examples)
+            hint = " — if that's a mm:ss clock, convert it to seconds" if bad_time_examples else ""
+            rep.add("W10", f"{rep._pct(len(bad_time))} have a start time that isn't a "
+                           f"number, so were left out (e.g. {eg}){hint}")
+        if blank_obs:
+            rep.add("W11", f"{rep._pct(len(blank_obs))} have a blank {observation} and "
+                           f"were left out (e.g. {_example_rows(blank_obs)})")
+        if blank_event:
+            rep.add("W12", f"{rep._pct(len(blank_event))} have a blank {event} and "
+                           f"were left out (e.g. {_example_rows(blank_event)})")
+        # W13: labels that differ only by case/spacing
+        groups: dict[str, set] = defaultdict(set)
+        for evs in rows_by_obs.values():
+            for _, code in evs:
+                groups[_norm_label(code)].add(code)
+        clashes = [sorted(v) for v in groups.values() if len(v) > 1]
+        if clashes:
+            eg = ", ".join(repr(x) for x in clashes[0])
+            more = f", and {len(clashes) - 1} more set(s)" if len(clashes) > 1 else ""
+            rep.add("W13", f"{len(clashes)} set(s) of labels differ only by "
+                           f"capitalisation or spacing (e.g. {eg}){more}; they are "
+                           f"counted as different events")
+        # W14: window problems
+        if bad_window or out_of_window:
+            bits = []
+            if bad_window:
+                bits.append(f"{bad_window} observation(s) have a window ending at or "
+                            f"before it starts")
+            if out_of_window:
+                bits.append(f"{out_of_window} event(s) fall outside their window")
+            rep.add("W14", "; ".join(bits))
+        # W15: time unit looks wrong
+        all_t = [t for evs in rows_by_obs.values() for t, _ in evs]
+        if all_t:
+            hi = max(all_t)
+            if time_unit == "ms" and hi < 1000 and len(all_t) > 5:
+                rep.add("W15", f"with time_unit='ms', every event falls in under one "
+                               f"second (largest time {hi} ms); the times may really be "
+                               f"in seconds")
+            elif time_unit == "s" and hi > 1_000_000:
+                rep.add("W15", f"with time_unit='s', the times run very large (up to "
+                               f"{hi} s, over 11 days); they may really be in milliseconds")
+
+        # I20: observations too small to host a pattern
+        tiny = sum(1 for o in sample if len(o.events) < 2)
+        if tiny:
+            rep.add("I20", f"{tiny} of {len(sample)} observations have fewer than two "
+                           f"events; they set the baseline rates but cannot contain a "
+                           f"pattern (the Methods text reports both counts)")
+        # I21: single observation
+        if len(sample) == 1:
+            rep.add("I21", "the file contains only one observation; patterns are "
+                           "counted across observations, so a single time-line will "
+                           "detect little")
+        # I22: a dominant event type
+        counts = Counter(code for evs in rows_by_obs.values() for _, code in evs)
+        total_ev = sum(counts.values())
+        if total_ev:
+            top, n_top = counts.most_common(1)[0]
+            if len(counts) > 1 and n_top / total_ev >= 0.30:
+                rep.add("I22", f"one event type, {top!r}, is {n_top / total_ev:.0%} of "
+                               f"all events; a very common event can link to almost "
+                               f"anything")
+        # I23: high co-timestamp fraction
+        adj = same = 0
+        for evs in rows_by_obs.values():
+            ts = sorted(t for t, _ in evs)
+            for a, b in zip(ts, ts[1:]):
+                adj += 1
+                if a == b:
+                    same += 1
+        if adj and same / adj >= 0.10:
+            rep.add("I23", f"{same / adj:.0%} of neighbouring events share an identical "
+                           f"time; their order is undefined at that resolution")
+
+    finish_or_raise()
+    return sample
+
+
+def check_reference() -> list[dict]:
+    """Return the check registry as ordered rows (code, severity, condition, fix).
+
+    Used to generate the tutorial's Appendix A from the same source the runtime
+    messages use, so the paper and the software cannot disagree.
+    """
+    label = {"E": "Error (stops the read)", "W": "Warning (loads anyway)",
+             "I": "Information (loads anyway)"}
+    return [{"code": c, "severity": label[m["severity"]],
+             "condition": m["condition"], "fix": m["fix"],
+             "example": m.get("example", "")}
+            for c, m in CHECKS.items()]
+
+
+def read_observation(path: str | Path) -> Observation:
+    """Read a single THEME ``.txt`` file into one :class:`Observation`.
+
+    THEME files have one ``time<TAB>event`` row per line, with ``:`` marking the
+    start of the window and ``&`` the end. Whitespace-separated files are tolerated,
+    and if a marker is missing the window falls back to the first and last event.
+    """
+    path = Path(path)
+    start: int | None = None
+    end: int | None = None
+    events: list[tuple[int, str]] = []
+
+    for raw in path.read_text().splitlines():
+        raw = raw.strip()
+        if not raw:
+            continue
+        parts = raw.split("\t")
+        if len(parts) != 2:
+            parts = raw.split()
+            if len(parts) != 2:
+                continue
+        tok_time, tok_event = parts[0].strip(), parts[1].strip()
+        if tok_time.lower() == "time":  # header row
+            continue
+        try:
+            t = int(tok_time)
+        except ValueError:
+            continue
+
+        if tok_event == START_MARKER:
+            start = t
+        elif tok_event == END_MARKER:
+            end = t
+        else:
+            events.append((t, tok_event))
+
+    events.sort(key=lambda e: e[0])
+    if start is None:
+        start = events[0][0] if events else 0
+    if end is None:
+        end = events[-1][0] if events else start
+    return Observation(name=path.stem, start=start, end=end, events=events)
+
+
+def read_sample(folder: str | Path, pattern: str = "*.txt") -> list[Observation]:
+    """Read every THEME file in a folder into a list of observations (a "sample")."""
+    folder = Path(folder)
+    files = sorted(folder.glob(pattern))
+    if not files:
+        raise FileNotFoundError(f"No files matching {pattern!r} in {folder}")
+    return [read_observation(f) for f in files]
+
+
+def read_sofcoder(path: str | Path, *, dataname: str = "DATANAME",
+                  t: str = "T", events: str = "Events", sep: str | None = None,
+                  code_sep: str = ",", frames_per_second: float | None = None,
+                  start_marker: str = ":", end_marker: str = "&") -> list[Observation]:
+    """Read a SOF-CODER / THEME data file (``DATANAME, T, Events``) into a sample.
+
+    SOF-CODER (Jonsson et al., 2006) writes one row per time frame, each row holding
+    several comma-separated codes that are active at that frame. This reader splits
+    each row so every code becomes one event at time ``T``; codes that share a frame
+    are co-timed, and ``min_lag`` / ``collapse_duplicates`` then decide how that
+    co-occurrence is treated. Each ``DATANAME`` becomes one :class:`Observation`, and
+    rows whose Events field is the ``:`` or ``&`` marker set the window (``T = end -
+    start`` is the NX/T denominator); without the markers the window falls back to
+    the first and last event.
+
+    The outer separator is auto-detected (tab if present, else whitespace) unless
+    ``sep`` is given; ``code_sep`` splits the Events field (default ``','``). ``T`` is
+    kept in the file's own frame unit unless ``frames_per_second`` is given, which
+    converts frames to milliseconds. A header naming the columns is used if present,
+    otherwise the columns are read positionally as (DATANAME, T, Events).
+
+    This is the way in for data collected in the SOF-CODER/THEME world; for a tool
+    that already exports one row per single event, use :func:`read_table`.
+    """
+    path = Path(path)
+    scale = (1000.0 / frames_per_second) if frames_per_second else 1.0
+
+    def split_row(line: str) -> list[str]:
+        if sep is not None:
+            parts = line.split(sep)
+        elif "\t" in line:
+            parts = line.split("\t")
+        else:
+            parts = line.split(None, 2)   # DATANAME, T, Events (Events holds no spaces)
+        return [p.strip() for p in parts]
+
+    with open(path, newline="") as fh:
+        lines = [ln.strip() for ln in fh if ln.strip()]
+    if not lines:
+        return []
+
+    header = [h.lower() for h in split_row(lines[0])]
+    if dataname.lower() in header and events.lower() in header:
+        di, ti, ei = (header.index(dataname.lower()), header.index(t.lower()),
+                      header.index(events.lower()))
+        body = lines[1:]
+    else:
+        di, ti, ei, body = 0, 1, 2, lines   # positional fallback
+
+    rows_by_obs: dict[str, list] = defaultdict(list)
+    starts: dict[str, int] = {}
+    ends: dict[str, int] = {}
+    for line in body:
+        parts = split_row(line)
+        if len(parts) <= max(di, ti, ei):
+            continue
+        obs_id = parts[di]
+        if not obs_id:
+            continue
+        try:
+            tv = int(round(float(parts[ti]) * scale))
+        except (ValueError, TypeError):
+            continue
+        raw = parts[ei]
+        if raw == start_marker:
+            starts.setdefault(obs_id, tv)
+            continue
+        if raw == end_marker:
+            ends[obs_id] = tv
+            continue
+        for code in raw.split(code_sep):
+            code = code.strip()
+            if code and code not in (start_marker, end_marker):
+                rows_by_obs[obs_id].append((tv, code))
+
+    sample: list[Observation] = []
+    for obs_id, evs in rows_by_obs.items():
+        evs.sort(key=lambda e: e[0])
+        o_start = starts.get(obs_id, evs[0][0])
+        o_end = ends.get(obs_id, evs[-1][0])
+        sample.append(Observation(name=obs_id, start=o_start, end=o_end, events=evs))
+    return sample

{tpattern-0.1.9 → tpattern-0.1.10}/tpattern/methods.py

@@ -48,6 +48,16 @@ def methods_text(config, *, observations=None, calibration=None,
         T = sum(max(o.end - o.start, 1) for o in observations)
         s.append(f"The sample comprised {n_obs} observations containing {n_ev} "
                  f"event occurrences over {T} time units of total observation window.")
+        # Report the effective sample: how many observations can actually host a
+        # pattern (two or more events). The rest still set the baseline rates, so
+        # they are not discarded, but reporting only the submitted count would
+        # overstate what was analysed.
+        n_host = sum(1 for o in observations if len(o.events) >= 2)
+        if n_host < n_obs:
+            s.append(f"Of these, {n_host} observations contained at least two events and "
+                     f"could therefore contain a pattern; the remaining {n_obs - n_host} "
+                     f"contributed to the baseline event rates but could not themselves "
+                     f"host a detected pattern.")
 
     # --- core detection ---
     s.append(

{tpattern-0.1.9 → tpattern-0.1.10/tpattern.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tpattern
-Version: 0.1.9
+Version: 0.1.10
 Summary: Open, auditable temporal & sequential pattern analysis (T-patterns) for sport and behaviour
 Author-email: Andrew Callaway <acallaway@bournemouth.ac.uk>
 License-Expression: MIT

{tpattern-0.1.9 → tpattern-0.1.10}/tpattern.egg-info/SOURCES.txt

@@ -6,6 +6,7 @@ tests/test_contrast.py
 tests/test_dedup.py
 tests/test_guided.py
 tests/test_null_compare.py
+tests/test_read_table_validation.py
 tests/test_report_dendrogram.py
 tests/test_reproduction.py
 tests/test_sofcoder.py

tpattern-0.1.9/tpattern/io.py

@@ -1,303 +0,0 @@
-"""
-Reading event data into a sample of observations.
-==================================================
-
-The primary entry point is :func:`read_table`, which reads the canonical flat CSV
-that any coding tool can export (one row per event: ``observation, event, start``;
-see SCHEMA.md). This is what almost all users want::
-
-    from tpattern import read_table
-    obs = read_table("events.csv")
-
-A T-pattern analysis runs over a *collection* of observations (a "sample"). Each
-observation is one independent time-line (a possession, rally, bout, match);
-patterns are counted across all observations, but a single pattern occurrence can
-never straddle two of them. Several events may share a timestamp — the detection
-engine treats co-timed events as an unordered set at that instant (see
-``Config.min_lag`` to require a genuine lag instead).
-
-:func:`read_sample` / :func:`read_observation` are a legacy reader for THEME's
-tab-separated observation files, kept for compatibility: one file per observation,
-``time<TAB>event`` rows, with ``:`` and ``&`` marking the observation window.
-"""
-
-from __future__ import annotations
-
-import csv
-from collections import defaultdict
-from dataclasses import dataclass, field
-from pathlib import Path
-
-
-# The two reserved symbols that delimit the observation window rather than
-# denoting a real event.
-START_MARKER = ":"
-END_MARKER = "&"
-
-
-@dataclass
-class Observation:
-    """One time-line: the events of a single observation file.
-
-    Attributes
-    ----------
-    name : str
-        File stem, used to trace patterns back to a specific shot/sequence.
-    start, end : int
-        The observation window [start, end] in milliseconds, taken from the
-        `:` and `&` markers. `T = end - start` is the window length used as the
-        denominator of the NX/T baseline probability.
-    events : list[tuple[int, str]]
-        (time, event_type) pairs, sorted by time, markers removed.
-    """
-
-    name: str
-    start: int
-    end: int
-    events: list[tuple[int, str]] = field(default_factory=list)
-
-    @property
-    def T(self) -> int:
-        """Length of the observation window in time units (ms)."""
-        return self.end - self.start
-
-
-def read_observation(path: str | Path) -> Observation:
-    """Parse a single THEME `.txt` file into an :class:`Observation`."""
-    path = Path(path)
-    start: int | None = None
-    end: int | None = None
-    events: list[tuple[int, str]] = []
-
-    for raw in path.read_text().splitlines():
-        raw = raw.strip()
-        if not raw:
-            continue
-        parts = raw.split("\t")
-        if len(parts) != 2:
-            # Tolerate whitespace-separated files as well.
-            parts = raw.split()
-            if len(parts) != 2:
-                continue
-        tok_time, tok_event = parts[0].strip(), parts[1].strip()
-        if tok_time.lower() == "time":  # header row
-            continue
-        try:
-            t = int(tok_time)
-        except ValueError:
-            continue
-
-        if tok_event == START_MARKER:
-            start = t
-        elif tok_event == END_MARKER:
-            end = t
-        else:
-            events.append((t, tok_event))
-
-    events.sort(key=lambda e: e[0])
-
-    # Fall back gracefully if a marker is missing: use the event span.
-    if start is None:
-        start = events[0][0] if events else 0
-    if end is None:
-        end = events[-1][0] if events else start
-
-    return Observation(name=path.stem, start=start, end=end, events=events)
-
-
-def read_sample(folder: str | Path, pattern: str = "*.txt") -> list[Observation]:
-    """Read every observation file in a folder into a list (a "sample")."""
-    folder = Path(folder)
-    files = sorted(folder.glob(pattern))
-    if not files:
-        raise FileNotFoundError(f"No files matching {pattern!r} in {folder}")
-    return [read_observation(f) for f in files]
-
-
-def read_table(path: str | Path, *, observation: str = "observation",
-               event: str = "event", start: str = "start",
-               end: str | None = "end", obs_start: str | None = "obs_start",
-               obs_end: str | None = "obs_end", build_event_from: list[str] | None = None,
-               time_unit: str = "s", sep: str = ",") -> list[Observation]:
-    """Read the canonical flat event table (see SCHEMA.md) into a sample.
-
-    One row per event; rows are grouped by the `observation` column into one
-    :class:`Observation` each. This is the software-agnostic entry point — any
-    tool (OpenTag, Sportscode, a spreadsheet) exports to these columns and the
-    rest of the library is unchanged.
-
-    Parameters
-    ----------
-    observation, event, start : str
-        Column names for the required fields.
-    end : str, optional
-        Column giving each event's end time; if absent the observation window is
-        taken from the min/max event `start`.
-    obs_start, obs_end : str, optional
-        Columns giving the *observation window* (constant within an observation).
-        These default to the conventional column names ("obs_start"/"obs_end"), so a
-        table that includes them is read correctly with **no arguments**; they are
-        ignored silently if the columns are absent. Supply a window when the unit has
-        real bounds that extend beyond its first and last event — a possession, rally
-        or bout usually does. The window length T = obs_end - obs_start is the
-        denominator of the NX/T baseline, so deriving it from the events instead will
-        shift every baseline probability. If absent, the window falls back to the
-        first/last event time.
-    build_event_from : list[str], optional
-        If given, the event code is built by joining these columns with '_'
-        (for coders who keep the code split across descriptor columns) instead of
-        reading `event` directly.
-    time_unit : {'s', 'ms'}
-        Unit of the time columns. Internally times are integer milliseconds, so
-        seconds are scaled by 1000 (the engine's baseline maths is unit-agnostic;
-        this only fixes the reported interval unit).
-    """
-    if time_unit not in ("s", "ms"):
-        raise ValueError(f"time_unit must be 's' or 'ms', got {time_unit!r}")
-    path = Path(path)
-    scale = 1000 if time_unit == "s" else 1
-    rows_by_obs: dict[str, list] = defaultdict(list)
-    ends_by_obs: dict[str, list] = defaultdict(list)
-    window_by_obs: dict[str, tuple] = {}
-
-    with open(path, newline="") as fh:
-        reader = csv.DictReader(fh, delimiter=sep)
-        cols = set(reader.fieldnames or [])
-        # the event code comes from `event`, unless it is built from other columns
-        code_cols = set(build_event_from) if build_event_from else {event}
-        required = {observation, start} | code_cols
-        missing = required - cols
-        if missing:
-            raise ValueError(
-                f"table {path.name} is missing required column(s) {sorted(missing)}. "
-                f"Found columns: {sorted(cols)}. Expected one row per event with at "
-                f"least '{observation}', '{event}' and '{start}' (rename via the "
-                f"read_table arguments if your export uses different names; see SCHEMA.md).")
-        for r in reader:
-            obs_id = (r.get(observation) or "").strip()
-            if not obs_id:
-                continue
-            if build_event_from:
-                # join only the non-empty parts, so an event with no value for one
-                # column (e.g. a Challenge with no outcome) yields "Challenge_Pressure",
-                # not "Challenge__Pressure", and one with none yields just its type.
-                parts = [(r.get(c) or "").strip() for c in build_event_from]
-                code = "_".join(p for p in parts if p)
-            else:
-                code = (r.get(event) or "").strip()
-            if not code:
-                continue
-            try:
-                t = int(round(float(r[start]) * scale))
-            except (ValueError, TypeError, KeyError):
-                continue
-            rows_by_obs[obs_id].append((t, code))
-            if end and r.get(end) not in (None, ""):
-                try:
-                    ends_by_obs[obs_id].append(int(round(float(r[end]) * scale)))
-                except ValueError:
-                    pass
-            if obs_start and obs_end and obs_id not in window_by_obs:
-                try:
-                    window_by_obs[obs_id] = (
-                        int(round(float(r[obs_start]) * scale)),
-                        int(round(float(r[obs_end]) * scale)))
-                except (ValueError, TypeError, KeyError):
-                    pass
-
-    sample: list[Observation] = []
-    for obs_id, evs in rows_by_obs.items():
-        evs.sort(key=lambda e: e[0])
-        if obs_id in window_by_obs:
-            o_start, o_end = window_by_obs[obs_id]
-        else:
-            o_start = evs[0][0]
-            o_end = max(ends_by_obs[obs_id]) if ends_by_obs.get(obs_id) else evs[-1][0]
-        sample.append(Observation(name=obs_id, start=o_start, end=o_end, events=evs))
-    return sample
-
-
-def read_sofcoder(path: str | Path, *, dataname: str = "DATANAME",
-                  t: str = "T", events: str = "Events", sep: str | None = None,
-                  code_sep: str = ",", frames_per_second: float | None = None,
-                  start_marker: str = ":", end_marker: str = "&") -> list[Observation]:
-    """Read a SOF-CODER / THEME data file (DATANAME, T, Events) into a sample.
-
-    SOF-CODER (Jonsson et al., 2006) exports one row per time frame, each row
-    carrying several comma-separated category codes active at that frame (its
-    multi-criteria "field format"). This reader explodes each row so every code
-    becomes one event at time ``T``; codes that share a frame are co-timed, and
-    ``min_lag``/``collapse_duplicates`` then decide how co-occurrence is treated.
-    Each ``DATANAME`` becomes one :class:`Observation`. Rows whose Events field is
-    the start (``:``) or end (``&``) marker set the observation window
-    ``[start, end]`` (``T = end - start`` is the NX/T baseline denominator); if the
-    markers are absent the window falls back to the first/last event.
-
-    The outer column delimiter is auto-detected (tab if present, else any
-    whitespace) unless ``sep`` is given; ``code_sep`` splits the Events field
-    (default ``','``). ``T`` is kept in the file's native frame unit unless
-    ``frames_per_second`` is supplied, in which case frames are converted to
-    milliseconds. A header row naming the columns is used if present, else the
-    columns are taken positionally as (DATANAME, T, Events).
-
-    This is the on-ramp for data collected in the SOF-CODER/THEME ecosystem; for
-    a coder that already exports one row per single event, use :func:`read_table`.
-    """
-    path = Path(path)
-    scale = (1000.0 / frames_per_second) if frames_per_second else 1.0
-
-    def split_row(line: str) -> list[str]:
-        if sep is not None:
-            parts = line.split(sep)
-        elif "\t" in line:
-            parts = line.split("\t")
-        else:
-            parts = line.split(None, 2)   # DATANAME, T, Events (Events holds no spaces)
-        return [p.strip() for p in parts]
-
-    with open(path, newline="") as fh:
-        lines = [ln.strip() for ln in fh if ln.strip()]
-    if not lines:
-        return []
-
-    header = [h.lower() for h in split_row(lines[0])]
-    if dataname.lower() in header and events.lower() in header:
-        di, ti, ei = (header.index(dataname.lower()), header.index(t.lower()),
-                      header.index(events.lower()))
-        body = lines[1:]
-    else:
-        di, ti, ei, body = 0, 1, 2, lines   # positional fallback
-
-    rows_by_obs: dict[str, list] = defaultdict(list)
-    starts: dict[str, int] = {}
-    ends: dict[str, int] = {}
-    for line in body:
-        parts = split_row(line)
-        if len(parts) <= max(di, ti, ei):
-            continue
-        obs_id = parts[di]
-        if not obs_id:
-            continue
-        try:
-            tv = int(round(float(parts[ti]) * scale))
-        except (ValueError, TypeError):
-            continue
-        raw = parts[ei]
-        if raw == start_marker:
-            starts.setdefault(obs_id, tv)
-            continue
-        if raw == end_marker:
-            ends[obs_id] = tv
-            continue
-        for code in raw.split(code_sep):
-            code = code.strip()
-            if code and code not in (start_marker, end_marker):
-                rows_by_obs[obs_id].append((tv, code))
-
-    sample: list[Observation] = []
-    for obs_id, evs in rows_by_obs.items():
-        evs.sort(key=lambda e: e[0])
-        o_start = starts.get(obs_id, evs[0][0])
-        o_end = ends.get(obs_id, evs[-1][0])
-        sample.append(Observation(name=obs_id, start=o_start, end=o_end, events=evs))
-    return sample