Hot start#

Note

Engine: OpenSWMM 6 — refactored.

Hot start files capture the hydraulic + quality state of a simulation at a moment in time so a follow-up run can resume from there.

Reference: openswmm_hotstart.h.

Save and apply#

from pathlib import Path
from openswmm.engine import Solver, HotStart

# ---- save ----
with Solver("warmup.inp") as s:
    for _ in s.steps():
        pass
    HotStart.save_from(s, Path("warmup.hs"))

# ---- apply ----
with HotStart.open("warmup.hs") as hs:
    print(hs.sim_datetime, hs.warnings)
    with Solver("storm.inp") as s2:
        # Apply in INITIALIZED state, before start().
        s2.open()
        s2.initialize()
        hs.apply(s2)
        s2.start()
        for _ in s2.steps():
            pass
        s2.end()

HotStart — lifecycle#

Operation

What it does

HotStart.open(path)

Classmethod. Opens an existing hot start file.

HotStart.save_from(solver, path)

Static method. Save current solver state to path.

hs.close() / with HotStart.open(p) as hs:

Context manager closes automatically.

hs.apply(solver)

Apply the captured state to a Solver in INITIALIZED state.

HotStart — metadata#

Property

Type

Meaning

sim_datetime

datetime

Moment the state was captured.

crs

str | None

CRS string from the file (or None).

warnings

list[str]

Warnings emitted during open / apply.

node_count / link_count

int

Counts inside the hot-start blob.

Direct state edits#

The C API stores hot-start state by id, not by index, so the editing methods are id-keyed:

with HotStart.open("warmup.hs") as hs:
    hs.set_node_depth("J1", 1.2)
    hs.set_node_head("J1", 102.3)
    hs.set_link_flow("C1", 0.4)
    hs.set_link_depth("C1", 0.6)
    hs.set_subcatchment_runoff("S1", 0.0)

Edits made on a HotStart are applied when apply() runs; they do not affect the on-disk file.

solver.save_schedule — recurring saves#

The engine’s [SAVE HOTSTART] block is exposed as a MutableSequence of SaveScheduleEntry records on the Solver:

from datetime import datetime
from openswmm.engine._hotstart import SaveScheduleEntry

solver.save_schedule.append(SaveScheduleEntry(
    when=datetime(2024, 6, 15, 12, 0),
    path="midday.hs",
))
for entry in solver.save_schedule:
    print(entry.when, entry.path)

del solver.save_schedule[0]
solver.save_schedule.clear()

Each entry is a NamedTuple so entry.when (a datetime) and entry.path (str) are stable attributes. The C API supports only append; insert is emulated by clearing and re-adding (just like solver.events).

Assignments accept either a SaveScheduleEntry, a (when, path) tuple, or a {"when": ..., "path": ...} dict.

See also#