Pandas HDFStore：当 min_itemsize 设置为字符串列的最大值时追加失败

Question

我正在检测多个数据帧的所有字符串列的最大长度，然后尝试构建 HDFStore：

import pandas as pd

# Detect max string length for each column across all DataFrames
max_lens = {}
for df_path in paths:
    df = pd.read_pickle(df_path)
    for col in df.columns:
        ser = df[col]
        if ser.dtype == 'object' and isinstance(
            ser.loc[ser.first_valid_index()], str
        ):
            max_lens[col] = max(
                ser.dropna().map(len).max(), max_lens.setdefault(col, 0)
            )
print('Setting min itemsizes:', max_lens)

hdf_path.unlink()  # Delete of file for clean retry
store = pd.HDFStore(hdf_path, complevel=9)
for df_path in paths:
    df = pd.read_pickle(df_path)
    store.append(hdf_key, df, min_itemsize=max_lens, data_columns=True)
store.close()

检测到的最大字符串长度如下：

     max_lens = {'hashtags': 139,
                 'id': 19,
                 'source': 157,
                 'text': 233,
                 'urls': 2352,
                 'user_mentions_user_ids': 199,
                 'in_reply_to_screen_name': 17,
                 'in_reply_to_status_id': 19,
                 'in_reply_to_user_id': 19,
                 'media': 286,
                 'place': 56,
                 'quoted_status_id': 19,
                 'user_id': 19}

但我仍然收到此错误：

ValueError: Trying to store a string with len [220] in [hashtags] column but
this column has a limit of [194]!
Consider using min_itemsize to preset the sizes on these columns

这很奇怪，因为检测到 hashtags 的最大长度是 139。

Answer 1

HDF以UTF-8存储字符串，因此需要将字符串编码为UTF-8，然后求最大长度。

a_pandas_string_series.str.encode('utf-8').str.len().max()

Pandas HDFStore：当 min_itemsize 设置为字符串列的最大值时追加失败

Pandas HDFStore: append fails when min_itemsize is set to the maximum of the string column

hdf5

pytables

pandas

hdfstore

python-3.7