将 pl.Duration 转换为人类字符串

Question

打印极坐标数据框时，默认情况下以“人类格式”打印

pl.Duration

。使用什么函数来执行此转换？可以使用吗？尝试

"{}".format()

返回一些可读但不那么好的内容。

import polars as pl

data = {"end": ["2024/11/13 10:28:00",
                "2024/10/10 10:10:10",
                "2024/09/13 09:12:29",
                "2024/08/31 14:57:02",
               ],
        "start": ["2024/11/13 10:27:33",
                  "2024/10/10 10:01:01",
                  "2024/09/13 07:07:07",
                  "2024/08/25 13:48:28",
        ]
       }

df = pl.DataFrame(data)
df = df.with_columns(
        pl.col("end").str.to_datetime(),
        pl.col("start").str.to_datetime(),
        )

df = df.with_columns(
        duration = pl.col("end") - pl.col("start"),
        )
df = df.with_columns(
        pl.col("duration").map_elements(lambda t: "{}".format(t), return_dtype=pl.String()).alias("duration_str")
        )
print(df)

shape: (4, 4)
┌─────────────────────┬─────────────────────┬──────────────┬─────────────────┐
│ end                 ┆ start               ┆ duration     ┆ duration_str    │
│ ---                 ┆ ---                 ┆ ---          ┆ ---             │
│ datetime[μs]        ┆ datetime[μs]        ┆ duration[μs] ┆ str             │
╞═════════════════════╪═════════════════════╪══════════════╪═════════════════╡
│ 2024-11-13 10:28:00 ┆ 2024-11-13 10:27:33 ┆ 27s          ┆ 0:00:27         │
│ 2024-10-10 10:10:10 ┆ 2024-10-10 10:01:01 ┆ 9m 9s        ┆ 0:09:09         │
│ 2024-09-13 09:12:29 ┆ 2024-09-13 07:07:07 ┆ 2h 5m 22s    ┆ 2:05:22         │
│ 2024-08-31 14:57:02 ┆ 2024-08-25 13:48:28 ┆ 6d 1h 8m 34s ┆ 6 days, 1:08:34 │
└─────────────────────┴─────────────────────┴──────────────┴─────────────────┘

Answer 1

有趣！！我不久前遇到了同样的问题，并首先应用了您的初始解决方案，但发现它并没有真正达到我想要的效果，我创建了一个函数来“模仿”持续时间格式。将其应用到您的数据作品中：

import polars as pl

data = {
    "end": ["2024/11/13 10:28:00", "2024/10/10 10:10:10", "2024/09/13 09:12:29", "2024/08/31 14:57:02"],
    "start": ["2024/11/13 10:27:33", "2024/10/10 10:01:01", "2024/09/13 07:07:07", "2024/08/25 13:48:28"],
}

df = pl.DataFrame(data)

df = df.with_columns(
    pl.col("end").str.strptime(pl.Datetime, format="%Y/%m/%d %H:%M:%S").alias("end"),
    pl.col("start").str.strptime(pl.Datetime, format="%Y/%m/%d %H:%M:%S").alias("start"),
)

df = df.with_columns(
    (pl.col("end") - pl.col("start")).alias("duration")
)

def format_duration(duration_microseconds):
    total_seconds = duration_microseconds // 1000000 
    days, rem = divmod(total_seconds, 86400)  
    hours, rem = divmod(rem, 3600)            
    minutes, seconds = divmod(rem, 60)
    parts = []
    if days > 0:
        parts.append(f"{days}d")
    if hours > 0 or days > 0:
        parts.append(f"{hours}h")
    if minutes > 0 or hours > 0 or days > 0:
        parts.append(f"{minutes}m")
    parts.append(f"{seconds}s")
    return " ".join(parts)

df = df.with_columns(
    pl.col("duration").cast(pl.Int64).map_elements(format_duration).alias("formatted_duration")
)

print(df)

并返回

shape: (4, 4)
┌─────────────────────┬─────────────────────┬──────────────┬────────────────────┐
│ end                 ┆ start               ┆ duration     ┆ formatted_duration │
│ ---                 ┆ ---                 ┆ ---          ┆ ---                │
│ datetime[μs]        ┆ datetime[μs]        ┆ duration[μs] ┆ str                │
╞═════════════════════╪═════════════════════╪══════════════╪════════════════════╡
│ 2024-11-13 10:28:00 ┆ 2024-11-13 10:27:33 ┆ 27s          ┆ 27s                │
│ 2024-10-10 10:10:10 ┆ 2024-10-10 10:01:01 ┆ 9m 9s        ┆ 9m 9s              │
│ 2024-09-13 09:12:29 ┆ 2024-09-13 07:07:07 ┆ 2h 5m 22s    ┆ 2h 5m 22s          │
│ 2024-08-31 14:57:02 ┆ 2024-08-25 13:48:28 ┆ 6d 1h 8m 34s ┆ 6d 1h 8m 34s       │
└─────────────────────┴─────────────────────┴──────────────┴────────────────────┘