Migrate from fitdecode¶

fitdecode is a clean frame-oriented reader, and its frame model maps almost one-to-one onto chiptime's lower layers — then chiptime's semantic layer covers the interpretation code you'd otherwise write on top.

API mapping¶

fitdecode	chiptime
`FitReader("a.fit")` + iterate frames	`chiptime.iter_frames("a.fit")` (wire) or `iter_messages` (decoded)
`frame.frame_type == FIT_FRAME_DATA`	messages from `iter_messages` are the data frames
`frame.name == "record"`	`msg.name == "record"`
`frame.get_value("hr", fallback=None)`	`msg.get("hr")` (absent → `None`)
`frame.get_field("hr").units`	`msg.fields["hr"].units`
`FitReader(..., check_crc=CrcCheck.WARN)`	default `lenient` (CRC triaged + reported)
`except FitHeaderError / FitEOFError:`	`result.errors` / `result.ok`; `strict` mode raises coded errors
your session/stream assembly code	`result.activity.sessions[*]` — already assembled, tested

Before / after¶

fitdecode

import fitdecode

power = []
with fitdecode.FitReader("ride.fit") as fit:
    for frame in fit:
        if frame.frame_type == fitdecode.FIT_FRAME_DATA and frame.name == "record":
            v = frame.get_value("power", fallback=None)
            if v is not None:
                power.append(v)

chiptime

import chiptime

result = chiptime.parse("ride.fit")
power = result.activity.sessions[0].records.stream("power")
# power.values: 0 W is coasting (real), None is dropout (absent) — kept distinct

What you can delete after migrating¶

Wrapper detection (gzip/zip) — chiptime unwraps and records it in source.unwrapped.
Sentinel guards — invalid values are None before you ever see them.
Session/lap/stream assembly, timer math, gap handling — the semantic layer ships them, with the edge cases (compressed timestamps, timestamp_16 rollover, HR expansion) already handled.