Solved

How to load sequences from a dataframe in batches?

  • 2 April 2024
  • 4 replies
  • 48 views

Hi everyone,

 

I'm loading sequences from a file that has 11 million rows, and I would like to load it in batches, where each batch is a dataframe. As I load by batch, the rows that the sequence already had are being deleted, leaving only the last batch. How can I load each batch without deleting the previous one?

 

This is the code I use to read the file in batches and load rows into the sequence. It is assumed that the sequence is already created.

 

Regards,

Karina Saylema

@Aditya Kotiyal  @HanishSharma  @Jason Dressel  @Jairo Salaya @Liliana Sierra 

 

 

 

icon

Best answer by Andrian Gasper 2 April 2024, 15:35

View original

4 replies

Userlevel 1
Badge +2

Hi @Karina Saylema 

Your code seems correct.
Could you make sure that “row_number” is unique? 

If you try the code bellow, are you encountering the same issue?
 

batch_size = 10000
external_id = 'TEST_UPLOAD4'
start_row = 0 # Initialize a starting row number

for chunk in pd.read_csv(file_input, sep=';', chunksize=batch_size):
chunk.drop(columns=["row_number"], inplace=True)
chunk.index = range(start_row, start_row + len(chunk))
client.sequences.data.insert_dataframe(chunk, external_id=external_id, dropna=False)
start_row += batch_size

 

Userlevel 4
Badge

@Karina Saylema 

I think what’s happening is that chunk.set_index() is resetting the index for each chunk which is resulting in the sequence rows being overwritten because the row indexes are the same as the prior batch.
I’m not a pandas expert, but something like explicitly setting the index based on your batch size might be what you are looking for: https://pandas.pydata.org/docs/reference/api/pandas.RangeIndex.html

Hope this helps,
Jason

Thanks @Andrian Gasper, @Jason Dressel for your help 

Userlevel 1
Badge +2

@Karina Saylema You can mark this questions as answered, if you believe it is. :) 

Reply