polars-ta 0.5.3__py3-none-any.whl → 0.5.5__py3-none-any.whl

polars_ta/wq/cross_sectional.py

@@ -17,7 +17,9 @@ _ols_kwargs = OLSKwargs(null_policy='drop', solve_method='svd')
 
 
 def cs_one_side(x: Expr, is_long: bool = True) -> Expr:
-    """Shifts all instruments up or down so that the Alpha becomes long-only or short-only
+    """横截面上，将全部资产上调或下调，使得 Alpha 策略转为纯多头配置（当方向参数设为空头时则转为纯空头配置）
+
+    Shifts all instruments up or down so that the Alpha becomes long-only or short-only
 (if side = short), respectively.
 
     Examples
@@ -54,7 +56,11 @@ def cs_one_side(x: Expr, is_long: bool = True) -> Expr:
 
 
 def cs_scale(x: Expr, scale_: float = 1, long_scale: float = 1, short_scale: float = 1) -> Expr:
-    """Scales input to booksize. We can also scale the long positions and short positions to separate scales by mentioning additional parameters to the operator.
+    """横截面上，将输入数据进行比例调整
+
+    此外，可通过向运算符添加额外参数，将多头头寸和空头头寸分别映射到独立的缩放比例上
+
+    Scales input to booksize. We can also scale the long positions and short positions to separate scales by mentioning additional parameters to the operator.
 
     Examples
     --------
@@ -95,7 +101,9 @@ def cs_scale(x: Expr, scale_: float = 1, long_scale: float = 1, short_scale: flo
 
 
 def cs_scale_down(x: Expr, constant: int = 0) -> Expr:
-    """Scales all values in each day proportionately between 0 and 1 such that minimum value maps to 0 and maximum value maps to 1.
+    """横截面上，将每日数据按比例缩放至 [0,1] 区间，使得最小值映射为 0，最大值映射为 1，并通过减去常数偏移量调整最终结果
+
+    Scales all values in each day proportionately between 0 and 1 such that minimum value maps to 0 and maximum value maps to 1.
     constant is the offset by which final result is subtracted
 
     Examples
@@ -130,7 +138,9 @@ def cs_scale_down(x: Expr, constant: int = 0) -> Expr:
 
 
 def cs_truncate(x: Expr, max_percent: float = 0.01) -> Expr:
-    """Operator truncates all values of x to maxPercent. Here, maxPercent is in decimal notation
+    """横截面上，将所有 x 的取值截断至 maxPercent 指定的上限值，其中 maxPercent 需以十进制小数形式表示
+
+    Operator truncates all values of x to maxPercent. Here, maxPercent is in decimal notation
 
     Examples
     --------
@@ -161,8 +171,8 @@ def cs_truncate(x: Expr, max_percent: float = 0.01) -> Expr:
     return x.clip(upper_bound=x.sum() * max_percent)
 
 
-def cs_fill_except_all_null(x: Expr, value=0) -> Expr:
-    """全为`null`时，保持`null`，反之`null`填充为`value`
+def cs_fill_except_all_null(x: Expr, value: float = 0) -> Expr:
+    """横截面上，全为`null`时，保持`null`，反之`null`填充为`value`
 
     Examples
     --------
@@ -199,27 +209,39 @@ def cs_fill_except_all_null(x: Expr, value=0) -> Expr:
 
 
 def cs_fill_mean(x: Expr) -> Expr:
-    """填充`null`为均值"""
+    """横截面上，填充`null`为均值"""
     return x.fill_null(strategy='mean')
 
 
-def cs_fill_null(x: Expr, value=0) -> Expr:
-    """填充`null`为`value`"""
+def cs_fill_max(x: Expr) -> Expr:
+    """横截面上，填充`null`为最大值"""
+    return x.fill_null(strategy='max')
+
+
+def cs_fill_min(x: Expr) -> Expr:
+    """横截面上，填充`null`为最小值"""
+    return x.fill_null(strategy='min')
+
+
+def cs_fill_null(x: Expr, value: float = 0) -> Expr:
+    """横截面上，填充`null`为`value`"""
     return x.fill_null(value)
 
 
 def cs_regression_neut(y: Expr, x: Expr) -> Expr:
-    """一元回归残差"""
+    """横截面上，一元回归残差"""
     return pls.compute_least_squares(y, x, add_intercept=True, mode='residuals', ols_kwargs=_ols_kwargs)
 
 
 def cs_regression_proj(y: Expr, x: Expr) -> Expr:
-    """一元回归预测"""
+    """横截面上，一元回归预测"""
     return pls.compute_least_squares(y, x, add_intercept=True, mode='predictions', ols_kwargs=_ols_kwargs)
 
 
 def cs_rank(x: Expr, pct: bool = True) -> Expr:
-    """排名。Ranks the input among all the instruments and returns an equally distributed number between 0.0 and 1.0. For precise sort, use the rate as 0.
+    """横截面排名
+
+    Ranks the input among all the instruments and returns an equally distributed number between 0.0 and 1.0. For precise sort, use the rate as 0.
 
     Parameters
     ----------
@@ -269,7 +291,7 @@ def cs_rank(x: Expr, pct: bool = True) -> Expr:
 
 
 def _cs_qcut_rank(x: Expr, q: int = 10) -> Expr:
-    """等频分箱
+    """横截面上等频分箱
 
     Parameters
     ----------
@@ -314,7 +336,9 @@ def _cs_qcut_rank(x: Expr, q: int = 10) -> Expr:
 
 
 def cs_qcut(x: Expr, q: int = 10) -> Expr:
-    """等频分箱 Convert float values into indexes for user-specified buckets. Bucket is useful for creating group values, which can be passed to group operators as input.
+    """横截面上等频分箱
+
+    Convert float values into indexes for user-specified buckets. Bucket is useful for creating group values, which can be passed to group operators as input.
 
     Parameters
     ----------
@@ -362,7 +386,7 @@ def cs_qcut(x: Expr, q: int = 10) -> Expr:
 
 
 def cs_top_bottom(x: Expr, k: int = 10) -> Expr:
-    """前K后K
+    """横截面上，排名。前K标记成-1，后K标记成1
 
     Examples
     --------

polars_ta/wq/logical.py

@@ -12,9 +12,9 @@ def equal(input1: Expr, input2: Expr) -> Expr:
     return input1 == input2
 
 
-def if_else(input1: Expr, input2: Expr, input3: Expr = None) -> Expr:
-    """If input1 is true then return input2 else return input3."""
-    return when(input1).then(input2).otherwise(input3)
+def if_else(condition: Expr, true_value: Expr, false_value: Expr = None) -> Expr:
+    """条件判断"""
+    return when(condition).then(true_value).otherwise(false_value)
 
 
 def is_finite(input1: Expr) -> Expr:

polars_ta/wq/preprocess.py

@@ -1,17 +1,28 @@
+"""
+补空值 → 去极值 → 标准化 → 中性化 → 标准化（可选二次标准化）
+
+# 对数市值。去极值
+MC_LOG = cs_quantile(log1p(market_cap), 0.01, 0.99)
+# 对数市值。标准化。供其他因子市值中性化时使用
+MC_NORM = cs_zscore(MC_LOG)
+# 对数市值。行业中性化。直接作为因子使用
+MC_NEUT = cs_zscore(cs_resid(MC_NORM, CS_SW_L1, ONE))
+
+"""
 import polars_ols as pls
 from polars import Expr, when
 from polars_ols.least_squares import OLSKwargs
 
-from polars_ta.wq.cross_sectional import cs_rank
-
 
 # ======================
 # standardize
 def cs_zscore(x: Expr, ddof: int = 0) -> Expr:
+    """横截面zscore标准化"""
     return (x - x.mean()) / x.std(ddof=ddof)
 
 
 def cs_minmax(x: Expr) -> Expr:
+    """横截面minmax标准化"""
     a = x.min()
     b = x.max()
     # 这个版本在b-a为整数时，得到的结果不好看
@@ -21,13 +32,15 @@ def cs_minmax(x: Expr) -> Expr:
 
 # ======================
 # winsorize
-def cs_quantile(x: Expr, low_limit: float = 0.025, up_limit: float = 0.995) -> Expr:
+def cs_quantile(x: Expr, low_limit: float = 0.025, up_limit: float = 0.975) -> Expr:
+    """横截面分位数去极值"""
     a = x.quantile(low_limit)
     b = x.quantile(up_limit)
     return x.clip(lower_bound=a, upper_bound=b)
 
 
 def cs_3sigma(x: Expr, n: float = 3.) -> Expr:
+    """横截面3倍sigma去极值"""
     # fill_nan will seriously reduce speed. So it's more appropriate for users to handle it themselves
     # fill_nan(None) 严重拖慢速度，所以还是由用户自己处理更合适
     a = x.mean()
@@ -36,7 +49,13 @@ def cs_3sigma(x: Expr, n: float = 3.) -> Expr:
 
 
 def cs_mad(x: Expr, n: float = 3., k: float = 1.4826) -> Expr:
-    # https://en.wikipedia.org/wiki/Median_absolute_deviation
+    """横截面MAD去极值
+
+    References
+    ----------
+    https://en.wikipedia.org/wiki/Median_absolute_deviation
+
+    """
     a = x.median()
     b = (n * k) * (x - a).abs().median()
     return x.clip(lower_bound=a - b, upper_bound=a + b)
@@ -45,7 +64,7 @@ def cs_mad(x: Expr, n: float = 3., k: float = 1.4826) -> Expr:
 # ======================
 # neutralize
 def cs_demean(x: Expr) -> Expr:
-    """demean
+    """横截面去均值化
 
     Notes
     -----
@@ -66,75 +85,26 @@ def cs_demean(x: Expr) -> Expr:
 _ols_kwargs = OLSKwargs(null_policy='drop', solve_method='svd')
 
 
-# def _residual_multiple(cols: List[Series], add_constant: bool) -> Series:
-#     # 将pl.Struct转成list,这样可以实现传正则，其它也转list
-#     cols = [list(c.struct) if isinstance(c.dtype, Struct) else [c] for c in cols]
-#     # 二维列表转一维列表，再转np.ndarray
-#     cols = [i.to_numpy() for p in cols for i in p]
-#     if add_constant:
-#         cols += [np.ones_like(cols[0])]
-#     yx = np.vstack(cols).T
-#
-#     # skip nan
-#     mask = np.any(np.isnan(yx), axis=1)
-#     yx_ = yx[~mask, :]
-#
-#     y = yx_[:, 0]
-#     x = yx_[:, 1:]
-#     coef = np.linalg.lstsq(x, y, rcond=None)[0]
-#     y_hat = np.sum(x * coef, axis=1)
-#     residual = y - y_hat
-#
-#     # refill
-#     out = np.empty_like(yx[:, 0])
-#     out[~mask] = residual
-#     out[mask] = np.nan
-#     return Series(out, nan_to_null=True)
-#
-#
-# def cs_resid_(y: Expr, *more_x: Expr) -> Expr:
-#     """multivariate regression
-#     多元回归
-#     """
-#     return map_batches([y, *more_x], lambda xx: _residual_multiple(xx, False))
-
-
 def cs_resid(y: Expr, *more_x: Expr) -> Expr:
-    """多元回归取残差"""
+    """横截面多元回归取残差"""
     return pls.compute_least_squares(y, *more_x, mode='residuals', ols_kwargs=_ols_kwargs)
 
 
+def cs_zscore_resid(y: Expr, *more_x: Expr) -> Expr:
+    """横截面标准化、中性化"""
+    return cs_resid(cs_zscore(y), *more_x)
+
+
 def cs_mad_zscore(y: Expr) -> Expr:
-    """去极值、标准化"""
+    """横截面去极值、标准化"""
     return cs_zscore(cs_mad(y))
 
 
 def cs_mad_zscore_resid(y: Expr, *more_x: Expr) -> Expr:
-    """去极值、标准化、中性化"""
+    """横截面去极值、标准化、中性化"""
     return cs_resid(cs_zscore(cs_mad(y)), *more_x)
 
 
-def cs_mad_rank(y: Expr) -> Expr:
-    """去极值，排名。"""
-    return cs_rank(cs_mad(y))
-
-
-def cs_mad_rank2(y: Expr, m: float) -> Expr:
-    """非线性处理。去极值，排名，移动峰或谷到零点，然后平方
-
-    适合于分层收益V型或倒V的情况"""
-    return (cs_rank(cs_mad(y)) - m) ** 2
-
-
-def cs_mad_rank2_resid(y: Expr, m: float, *more_x: Expr) -> Expr:
-    """非线性处理。去极值，排名，移动峰或谷到零点，然后平方。回归取残差
-
-    适合于分层收益V型或倒V的情况"""
-    return cs_resid((cs_rank(cs_mad(y)) - m) ** 2, *more_x)
-
-
-def cs_rank2(y: Expr, m: float) -> Expr:
-    """非线性处理。移动峰或谷到零点，然后平方
-
-    适合于分层收益V型或倒V的情况"""
-    return (cs_rank(y) - m) ** 2
+def cs_mad_zscore_resid_zscore(y: Expr, *more_x: Expr) -> Expr:
+    """横截面去极值、标准化、中性化、二次标准化"""
+    return cs_zscore(cs_resid(cs_zscore(cs_mad(y)), *more_x))

polars_ta/wq/time_series.py

@@ -12,7 +12,11 @@ from polars_ta.wq._nb import roll_argmax, roll_argmin, roll_co_kurtosis, roll_co
 
 
 def ts_arg_max(x: Expr, d: int = 5, reverse: bool = True, min_samples: Optional[int] = None) -> Expr:
-    """Returns the relative index of the max value in the time series for the past d days.
+    """最大值相对位置
+
+    最近的一天记为第 0 天，最远的一天为第 d-1 天
+
+    Returns the relative index of the max value in the time series for the past d days.
     If the current day has the max value for the past d days, it returns 0.
     If previous day has the max value for the past d days, it returns 1.
 
@@ -61,7 +65,9 @@ def ts_arg_max(x: Expr, d: int = 5, reverse: bool = True, min_samples: Optional[
 
 
 def ts_arg_min(x: Expr, d: int = 5, reverse: bool = True, min_samples: Optional[int] = None) -> Expr:
-    """
+    """最小值相对位置
+
+    最近的一天记为第 0 天，最远的一天为第 d-1 天
 
     Parameters
     ----------
@@ -85,19 +91,21 @@ def ts_arg_min(x: Expr, d: int = 5, reverse: bool = True, min_samples: Optional[
 
 
 def ts_co_kurtosis(x: Expr, y: Expr, d: int = 5, ddof: int = 0, min_samples: Optional[int] = None) -> Expr:
+    """计算两个序列在滚动窗口内联合分布的协峰度"""
     minp = min_samples or polars_ta.MIN_SAMPLES or d
     return struct([x, y]).map_batches(lambda xx: batches_i2_o1(struct_to_numpy(xx, 2), roll_co_kurtosis, d, minp))
 
 
 def ts_co_skewness(x: Expr, y: Expr, d: int = 5, ddof: int = 0, min_samples: Optional[int] = None) -> Expr:
+    """计算两个序列在滚动窗口内联合分布的协偏度"""
     minp = min_samples or polars_ta.MIN_SAMPLES or d
     return struct([x, y]).map_batches(lambda xx: batches_i2_o1(struct_to_numpy(xx, 2), roll_co_skewness, d, minp))
 
 
 def ts_corr(x: Expr, y: Expr, d: int = 5, ddof: int = 1, min_samples: Optional[int] = None) -> Expr:
-    """rolling correlation between two columns
+    """时序滚动相关系数
 
-    时序滚动相关系数
+    rolling correlation between two columns
 
     Parameters
     ----------
@@ -236,9 +244,9 @@ def ts_count_nulls(x: Expr, d: int = 5, min_samples: Optional[int] = None) -> Ex
 
 
 def ts_covariance(x: Expr, y: Expr, d: int = 5, ddof: int = 1, min_samples: Optional[int] = None) -> Expr:
-    """rolling covariance between two columns
+    """时序滚动协方差
 
-    时序协方差
+    rolling covariance between two columns
 
     Parameters
     ----------
@@ -291,7 +299,7 @@ def ts_cum_count(x: Expr) -> Expr:
 
 
 def ts_cum_max(x: Expr) -> Expr:
-    """时序累计最大
+    """时序累计最大值
 
     Examples
     --------
@@ -323,7 +331,7 @@ def ts_cum_max(x: Expr) -> Expr:
 
 
 def ts_cum_min(x: Expr) -> Expr:
-    """时序累计最小
+    """时序累计最小值
 
     Examples
     --------
@@ -492,8 +500,10 @@ def ts_decay_linear(x: Expr, d: int = 30, min_samples: Optional[int] = None) ->
     return x.map_batches(lambda x1: batches_i1_o1(x1.to_numpy().astype(float), roll_decay_linear, d, minp))
 
 
-def ts_delay(x: Expr, d: int = 1, fill_value=None) -> Expr:
-    """时序数据移动 shift x
+def ts_delay(x: Expr, d: int = 1, fill_value: float = None) -> Expr:
+    """时序数据移动
+
+    shift x
 
     Parameters
     ----------
@@ -508,7 +518,7 @@ def ts_delay(x: Expr, d: int = 1, fill_value=None) -> Expr:
 
 
 def ts_delta(x: Expr, d: int = 1) -> Expr:
-    """差分"""
+    """时序差分"""
     return x.diff(d)
 
 
@@ -551,14 +561,16 @@ def ts_fill_null(x: Expr, limit: int = None) -> Expr:
 
 
 def ts_ir(x: Expr, d: int = 1, min_samples: Optional[int] = None) -> Expr:
-    """时序滚动信息系数rolling information ratio"""
+    """时序滚动信息系数
+
+    rolling information ratio"""
     return ts_mean(x, d, min_samples) / ts_std_dev(x, d, 0, min_samples)
 
 
 def ts_kurtosis(x: Expr, d: int = 5, bias: bool = False, min_samples: Optional[int] = None) -> Expr:
-    """kurtosis of x for the last d days
+    """时序滚动峰度
 
-    时序滚动峰度
+    kurtosis of x for the last d days
 
     Parameters
     ----------
@@ -603,16 +615,18 @@ def ts_kurtosis(x: Expr, d: int = 5, bias: bool = False, min_samples: Optional[i
 
 
 def ts_l2_norm(x: Expr, d: int = 5, min_samples: Optional[int] = None) -> Expr:
-    """Euclidean norm
+    """欧几里得范数
 
-    欧几里得范数"""
+    Euclidean norm
+    """
     minp = min_samples or polars_ta.MIN_SAMPLES
     return x.pow(2).rolling_sum(d, min_samples=minp).sqrt()
 
 
 def ts_log_diff(x: Expr, d: int = 1) -> Expr:
-    """对数差分。log(current value of input or x[t] ) - log(previous value of input or x[t-1]).
+    """求对数，然后时序滚动差分
 
+    log(current value of input or x[t] ) - log(previous value of input or x[t-1]).
     """
     return x.log().diff(d)
 
@@ -624,7 +638,9 @@ def ts_max(x: Expr, d: int = 30, min_samples: Optional[int] = None) -> Expr:
 
 
 def ts_max_diff(x: Expr, d: int = 30, min_samples: Optional[int] = None) -> Expr:
-    """Returns x - ts_max(x, d)"""
+    """窗口内最大值与当前值的差异‌
+
+    x - ts_max(x, d)"""
     return x - ts_max(x, d, min_samples)
 
 
@@ -647,24 +663,30 @@ def ts_min(x: Expr, d: int = 30, min_samples: Optional[int] = None) -> Expr:
 
 
 def ts_min_diff(x: Expr, d: int = 30, min_samples: Optional[int] = None) -> Expr:
-    """Returns x - ts_min(x, d)"""
+    """窗口内最小值与当前值的差异‌
+
+    x - ts_min(x, d)"""
     return x - ts_min(x, d, min_samples)
 
 
 def ts_min_max_cps(x: Expr, d: int, f: float = 2.0, min_samples: Optional[int] = None) -> Expr:
-    """Returns (ts_min(x, d) + ts_max(x, d)) - f * x"""
+    """计算时间窗口内最小值与最大值的总和减去当前值的加权结果
+
+    (ts_min(x, d) + ts_max(x, d)) - f * x"""
     return (ts_min(x, d, min_samples) + ts_max(x, d, min_samples)) - f * x
 
 
 def ts_min_max_diff(x: Expr, d: int, f: float = 0.5, min_samples: Optional[int] = None) -> Expr:
-    """Returns x - f * (ts_min(x, d) + ts_max(x, d))"""
+    """计算当前值 x 与基于时间窗口内最小值、最大值的加权组合的差值
+
+    x - f * (ts_min(x, d) + ts_max(x, d))"""
     return x - f * (ts_min(x, d, min_samples) + ts_max(x, d, min_samples))
 
 
 def ts_moment(x: Expr, d: int, k: int = 0, min_samples: Optional[int] = None) -> Expr:
-    """Returns K-th central moment of x for the past d days.
+    """滚动k阶中心距
 
-    滚动k阶中心距
+    Returns K-th central moment of x for the past d days.
 
     Parameters
     ----------
@@ -679,18 +701,19 @@ def ts_moment(x: Expr, d: int, k: int = 0, min_samples: Optional[int] = None) ->
 
 
 def ts_partial_corr(x: Expr, y: Expr, z: Expr, d: int, min_samples: Optional[int] = None) -> Expr:
-    """Returns partial correlation of x, y, z for the past d days.
+    """滚动偏相关
+
+    Returns partial correlation of x, y, z for the past d days.
 
-    滚动偏相关
     """
     minp = min_samples or polars_ta.MIN_SAMPLES or d
     return struct([x, y, z]).map_batches(lambda xx: batches_i2_o1(struct_to_numpy(xx, 3), roll_partial_corr, d, minp))
 
 
 def ts_percentage(x: Expr, d: int, percentage: float = 0.5, min_samples: Optional[int] = None) -> Expr:
-    """Returns percentile value of x for the past d days.
+    """滚动百分位数
 
-    滚动百分位数
+    Returns percentile value of x for the past d days.
 
     Parameters
     ----------
@@ -722,15 +745,22 @@ def ts_rank(x: Expr, d: int = 5, min_samples: Optional[int] = None) -> Expr:
     return x.map_batches(lambda a: roll_rank(a, d, minp, True))
 
 
+def ts_realized_volatility(close: Expr, d: int = 5, min_samples: Optional[int] = None) -> Expr:
+    """已实现波动率"""
+    minp = min_samples or polars_ta.MIN_SAMPLES or d
+    return ts_log_diff(close, 1).rolling_std(d, ddof=0, min_samples=minp)
+
+
 def ts_returns(x: Expr, d: int = 1) -> Expr:
     """简单收益率"""
     return x.pct_change(d)
 
 
 def ts_scale(x: Expr, d: int = 5, min_samples: Optional[int] = None) -> Expr:
-    """Returns (x – ts_min(x, d)) / (ts_max(x, d) – ts_min(x, d)) + constant
+    """时序滚动缩放。相当于ts_minmax
+
+    Returns (x – ts_min(x, d)) / (ts_max(x, d) – ts_min(x, d)) + constant
 
-    时序滚动缩放
     """
     a = ts_min(x, d, min_samples)
     b = ts_max(x, d, min_samples)
@@ -739,9 +769,9 @@ def ts_scale(x: Expr, d: int = 5, min_samples: Optional[int] = None) -> Expr:
 
 
 def ts_skewness(x: Expr, d: int = 5, bias: bool = False, min_samples: Optional[int] = None) -> Expr:
-    """Return skewness of x for the past d days
+    """时序滚动偏度
 
-    时序滚动偏度
+    Return skewness of x for the past d days
 
     Parameters
     ----------
@@ -836,8 +866,9 @@ def ts_sum_split_by(x: Expr, by: Expr, d: int = 30, k: int = 10) -> Expr:
 
 
 def ts_triple_corr(x: Expr, y: Expr, z: Expr, d: int, min_samples: Optional[int] = None) -> Expr:
-    """时序滚动三重相关系数 Returns triple correlation of x, y, z for the past d days.
+    """时序滚动三重相关系数
 
+    Returns triple correlation of x, y, z for the past d days.
 
     """
     minp = min_samples or polars_ta.MIN_SAMPLES or d
@@ -845,7 +876,9 @@ def ts_triple_corr(x: Expr, y: Expr, z: Expr, d: int, min_samples: Optional[int]
 
 
 def ts_weighted_decay(x: Expr, k: float = 0.5, min_samples: Optional[int] = None) -> Expr:
-    """加权衰减 Instead of replacing today’s value with yesterday’s as in ts_delay(x, 1),
+    """时序滚动加权衰减求和
+
+    Instead of replacing today’s value with yesterday’s as in ts_delay(x, 1),
     it assigns weighted average of today’s and yesterday’s values with weight on today’s value being k and yesterday’s being (1-k).
 
     Parameters

polars_ta/wq/transformational.py

@@ -1,4 +1,4 @@
-from polars import Expr, when, Boolean, Int32
+from polars import Expr, when, Boolean, Int32, Float32
 
 
 def cut(x: Expr, b: float, *more_bins) -> Expr:
@@ -86,7 +86,7 @@ def clamp(x: Expr, lower: float = 0, upper: float = 0, inverse: bool = False, ma
 #     raise
 
 
-def keep(x: Expr, f: float, period: int = 5) -> Expr:
+def _keep(x: Expr, f: float, period: int = 5) -> Expr:
     """This operator outputs value x when f changes and continues to do that for “period” days after f stopped changing. After “period” days since last change of f, NaN is output."""
     raise
 
@@ -211,7 +211,7 @@ def right_tail(x: Expr, minimum: float = 0) -> Expr:
 
 
 def sigmoid(x: Expr) -> Expr:
-    """Returns 1 / (1 + exp(-x))"""
+    """sigmoid激活函数"""
     return 1 / (1 + (-x).exp())
 
 
@@ -258,10 +258,15 @@ def tail(x: Expr, lower: float = 0, upper: float = 0, newval: float = 0) -> Expr
 
 
 def int_(a: Expr) -> Expr:
-    """convert bool to int"""
+    """bool转int"""
     return a.cast(Int32)
 
 
 def bool_(a: Expr) -> Expr:
-    """convert int to bool"""
+    """int转成bool"""
     return a.cast(Boolean)
+
+
+def float_(a: Expr) -> Expr:
+    """int转成float"""
+    return a.cast(Float32)

{polars_ta-0.5.3.dist-info → polars_ta-0.5.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: polars_ta
-Version: 0.5.3
+Version: 0.5.5
 Summary: polars expressions
 Author-email: wukan <wu-kan@163.com>
 License: MIT License
@@ -24,21 +24,19 @@ License: MIT License
         LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
         OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
         SOFTWARE.
-        
-Keywords: polars,expression,talib
+License-File: LICENSE
+Keywords: expression,polars,talib
 Classifier: Development Status :: 4 - Beta
 Classifier: Programming Language :: Python
 Requires-Python: >=3.8
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: polars>=1.28.0
-Requires-Dist: polars-ols>=0.3.0
-Requires-Dist: numpy
 Requires-Dist: numba
+Requires-Dist: numpy
 Requires-Dist: pandas
+Requires-Dist: polars-ols>=0.3.0
+Requires-Dist: polars>=1.28.0
 Provides-Extra: talib
-Requires-Dist: TA-Lib; extra == "talib"
-Dynamic: license-file
+Requires-Dist: ta-lib; extra == 'talib'
+Description-Content-Type: text/markdown
 
 # polars_ta
 
@@ -93,7 +91,7 @@ df = df.with_columns([
 ])
 ```
 
-When both `min_periods` and `MIN_SAMPLES` are set, `min_periods` takes precedence. default value is `None`.
+When both `min_samples` and `MIN_SAMPLES` are set, `min_samples` takes precedence. default value is `None`.
 
 ```python
 import polars_ta
@@ -196,7 +194,7 @@ df = df.with_columns([
 ])
 ```
 
-当`min_periods`和`MIN_SAMPLES`都设置时，以`min_periods`为准，默认值为`None`
+当`min_samples`和`MIN_SAMPLES`都设置时，以`min_samples`为准，默认值为`None`
 
 ```python
 import polars_ta
@@ -248,6 +246,11 @@ mkdocs build
 也可以通过以下链接导入：
 https://polars-ta.readthedocs.io/en/latest/llms-full.txt
 
+## 提示词
+由于`llms-full.txt`信息不适合做提示词，所以`tools/prompt.py`提供了生成更简洁算子清单的功能。
+
+用户也可以直接使用`prompt.txt`(欢迎提示词工程专家帮忙改进，做的更准确)
+
 ## 参考
 
 - https://github.com/pola-rs/polars