Communities

Writing
Writing
Codidact Meta
Codidact Meta
The Great Outdoors
The Great Outdoors
Photography & Video
Photography & Video
Scientific Speculation
Scientific Speculation
Cooking
Cooking
Electrical Engineering
Electrical Engineering
Judaism
Judaism
Languages & Linguistics
Languages & Linguistics
Software Development
Software Development
Mathematics
Mathematics
Christianity
Christianity
Code Golf
Code Golf
Music
Music
Physics
Physics
Linux Systems
Linux Systems
Power Users
Power Users
Tabletop RPGs
Tabletop RPGs
Community Proposals
Community Proposals
tag:snake search within a tag
answers:0 unanswered questions
user:xxxx search by author id
score:0.5 posts with 0.5+ score
"snake oil" exact phrase
votes:4 posts with 4+ votes
created:<1w created < 1 week ago
post_type:xxxx type of post
Search help
Notifications
Mark all as read See all your notifications »
Q&A

Welcome to Software Development on Codidact!

Will you help us build our independent community of developers helping developers? We're small and trying to grow. We welcome questions about all aspects of software development, from design to code to QA and more. Got questions? Got answers? Got code you'd like someone to review? Please join us.

Comments on How to get conditional running cumulative sum based on current row and previous rows?

Post

How to get conditional running cumulative sum based on current row and previous rows?

+0
−1

How do I perform a running cumulative sum that is based on a condition involving the current row and previous rows?

Given the following table:

acc | value | threshold
3   | 1     | 1
1   | 2     | 2
2   | 3     | 2

I would like to find the cumulative sum of acc if value >= threshold, for all values from the start to the current row. The expected output should be 3, 1, 3.

That is, the equivalent python code might look like:

for i in len(df):
    for j in range(i):
        if df[j].value >= df[i].threshold:
            df[i].cumsum += df[j].value

I tried using a windowed sum:

import pyspark.sql.functions as F
from pyspark.sql.window import Window

df = spark.createDataFrame([(3, 1, 1), (1, 2, 2), (2, 3, 2)], ["acc", "value", "threshold"])
window = Window.rowsBetween(Window.unboundedPreceding, Window.currentRow)
display(df.withColumn("output", F.sum(F.when(F.col("value") >= F.col("threshold"), F.col("acc"))).over(window)))

But this gave 3, 4, 6, because it was comparing against the same threshold on each row.

History
Why does this post require attention from curators or moderators?
You might want to add some details to your flag.
Why should this post be closed?

3 comment threads

Discussed on meta (1 comment)
Should be closed (2 comments)
Two questions in one (6 comments)
Two questions in one
matthewsnyder‭ wrote 5 months ago

It would be more useful to split this into two questions:

  1. How to filter rows based on a condition
  2. How to do a cumsum

Combining them makes things more confusing, especially to people who want to know only one and not the other.

matthewsnyder‭ wrote 5 months ago

Also, the following code is simpler:

# Part 1: Filter
filtered = [i for i in all_data if i["value"] > i["threshold"]]

# Part 2: Cumsum
s = 0
for i in filtered:
      s += i["value"]
      i["cumsum"] = s
congusbongus‭ wrote 5 months ago

you have misinterpreted the question; the condition for the cumsum involves different rows for "value" and "threshold", therefore the filter and cumsum cannot be split into two parts, as it would lead to the incorrect result of "3, 4, 6" instead of "3, 1, 3" as expected. Further, and this may not have been your intention, but the question specifies apache-spark/pyspark so plain python code cannot be the answer as it would be inefficient, although it can be used to clarify the question.

Skipping 1 deleted comment.

matthewsnyder‭ wrote 5 months ago

Oh, I see now. You're not taking a cumulative sum, you're taking a sum of all values below each threshold. So there's a self-join.

You're right, my Spark-less code won't work. The one you have in the question will.

You should remove mentions of "cumulative sum" because that's not what you are trying to do.

matthewsnyder‭ wrote 5 months ago

Still 2 questions btw:

  1. Self-join all values that are more than threshold.
  2. Sum them (with a group by).
congusbongus‭ wrote 5 months ago

I don't think it's necessarily those two questions, since there are alternate solutions that don't involve a self-join and sum (for example, 2 nested for loops, although possibly less efficient, would be a solution too).