集合数据、文件与错误处理

要求输入是一列非空文本编号，输出保留每个编号第一次出现的顺序。集合负责成员查询，列表负责顺序。

def unique_ids(ids):
    seen = set()
    result = []

    for position, item in enumerate(ids):
        if not is_nonempty_text(item):
            raise ValueError("invalid id at position " + str(position))
        if item not in seen:
            seen.add(item)
            result.append(item)

    return result

输入 ["b", "a", "b", "c"] 返回 ["b", "a", "c"]；空序列返回空序列；含空文本时立即失败并报告位置。若只有集合而直接输出其遍历结果，接口就没有明确保留首次顺序的保证。平均情况下循环是 $O(n)$，额外空间至多存放全部不同编号。

文件契约要求三列 sample_id、time_s、temperature_K。编号非空且唯一，时间有限且非负，温度有限且不低于零开尔文。只有全部记录通过后才返回。

def read_temperatures(path):
    rows = []
    seen_ids = set()

    try:
        with open(path, "r", encoding="utf-8", newline="") as handle:
            reader = csv_dict_reader(handle)
            require_exact_columns(
                reader.fieldnames,
                ["sample_id", "time_s", "temperature_K"],
            )

            for line_number, raw in enumerate(reader, start=2):
                sample_id = raw["sample_id"].strip()
                if sample_id == "":
                    raise DataError("empty sample_id", line_number)
                if sample_id in seen_ids:
                    raise DataError("duplicate sample_id", line_number)

                time_s = parse_finite_float(raw["time_s"], line_number)
                temperature_K = parse_finite_float(
                    raw["temperature_K"], line_number
                )
                if time_s < 0:
                    raise DataError("negative time", line_number)
                if temperature_K < 0:
                    raise DataError("temperature below 0 K", line_number)

                seen_ids.add(sample_id)
                rows.append((sample_id, time_s, temperature_K))
    except UnicodeError as error:
        raise InputError("file is not valid UTF-8: " + str(path)) from error
    except CsvSyntaxError as error:
        raise InputError("invalid CSV in " + str(path)) from error
    except OSError as error:
        raise InputError("cannot read " + str(path)) from error

    return rows

若文件不存在，外层异常保留原始系统原因；若第三行温度为 "hot"，转换函数报告行号；若第十行重复编号，函数失败而不返回前九行的“部分成功”。上下文块在所有路径关闭文件。是否允许额外列是模式政策，本例选择严格拒绝以发现拼写错误。

约定配置只含 time_step_s、steps、output_path。时间步是有限正数，步数是正整数，路径是非空文本。

def load_config(path):
    try:
        with open(path, "r", encoding="utf-8") as handle:
            raw = json_load(handle)
    except JsonSyntaxError as error:
        raise InputError("invalid JSON in " + str(path)) from error
    except UnicodeError as error:
        raise InputError("file is not valid UTF-8: " + str(path)) from error
    except OSError as error:
        raise InputError("cannot read " + str(path)) from error

    if not is_mapping(raw):
        raise SchemaError("top level must be an object")

    require_exact_keys(raw, {"time_step_s", "steps", "output_path"})
    dt = raw["time_step_s"]
    steps = raw["steps"]
    output_path = raw["output_path"]

    if not is_finite_number(dt) or dt <= 0:
        raise SchemaError("time_step_s must be positive and finite")
    if not is_integer(steps) or steps < 1:
        raise SchemaError("steps must be a positive integer")
    if not is_nonempty_text(output_path):
        raise SchemaError("output_path must be nonempty text")

    return Config(dt, steps, output_path)

缺失 steps、把 steps 写成文本 "1000"、给时间步 null 或增加拼错字段 time_steps_s 都会失败。严格模式避免程序悄悄使用默认值跑出另一组实验。若产品需要向后兼容，应给模式版本和显式迁移，而不是对所有未知字段静默忽略。

本例中的 is_integer 应明确排除布尔值，因为“真假开关”和“步数”在业务语义上不同，即使某些语言允许它们参与相同算术。验证函数应检查领域类型，而不只检查值能否被乘除。

输入记录为实验组名和有限测量值。先累计总和与计数，再构造新映射：

def group_means(records):
    totals = {}
    counts = {}

    for position, record in enumerate(records):
        group, value = validate_record(record, position)
        totals[group] = totals.get(group, 0.0) + value
        counts[group] = counts.get(group, 0) + 1

    result = {}
    for group in totals:
        result[group] = totals[group] / counts[group]
    return result

空输入返回空映射，不发生除零，因为没有创建空组。非法记录在加入累计前失败，避免总和已改而计数未改。若需要稳定输出顺序，写文件时显式排序组名；计算映射本身只表达组到平均值的关系。