非周期性数据值移位处理问题：pandas.Series.shift()无法满足需求

Looking at your data and desired output, what you need is to fill prev_value with the value of the previous consecutive group from new_value—with the first group using its own value as the prev_value. Here's a straightforward way to do this with pandas, even without periodic data:

Step-by-Step Implementation #

• Convert your data to a pandas DataFrame
  First, let's get your raw data into a DataFrame for easier manipulation:

  import pandas as pd
  
  data = {
      'new_value': ['100', '100', '250', '250', '250', '50', '90', '90', '90', '90'],
      'prev_value': ['None'] * 10
  }
  df = pd.DataFrame(data)
  

• Identify consecutive groups in new_value
  We can create a group identifier by checking where the value of new_value changes from the previous row. Using a shift comparison will help us spot these changes, then we can cumulative sum to assign unique group IDs:

  # Create a boolean mask where new_value changes from the prior row
  group_mask = df['new_value'] != df['new_value'].shift(1)
  # Assign unique group IDs starting from 0
  df['group_id'] = group_mask.cumsum() - 1
  

• Map each group to its corresponding prev_value
  Now, we'll create a lookup dictionary that maps each group ID to its appropriate prev_value:

  • For the first group (ID 0), prev_value is the group's own new_value
  • For subsequent groups, prev_value is the new_value of the previous group

  # Get the unique value for each group
  group_values = df.groupby('group_id')['new_value'].first().reset_index()
  # Shift group values to get the previous group's value
  group_values['prev_group_value'] = group_values['new_value'].shift(1)
  # Fill the first group's prev with its own value
  group_values.loc[0, 'prev_group_value'] = group_values.loc[0, 'new_value']
  # Convert to a dictionary for quick lookup
  prev_lookup = group_values.set_index('group_id')['prev_group_value'].to_dict()
  

• Fill the prev_value column using the lookup
  Finally, we map the group_id to the lookup dictionary to populate prev_value:

  df['prev_value'] = df['group_id'].map(prev_lookup)
  # Optional: drop the group_id column if you don't need it
  df = df.drop('group_id', axis=1)
  

Verify the Result #

If you print out df, you'll get exactly your desired output:

new_value prev_value
0       100        100
1       100        100
2       250        100
3       250        100
4       250        100
5        50        250
6        90         50
7        90         50
8        90         50
9        90         50

To convert it back to the dictionary format you want:

exp_result = df.to_dict('list')

This approach works regardless of how many consecutive values are in each group, and doesn't rely on periodicity—perfect for your use case!

内容的提问来源于stack exchange，提问作者savchart