如何计算 pandas 中分组行之间的差异

Question

I have a dataset with the number of views per article.

I'm trying to calculate the additional number of views per day, for each story, so I can graph it.

I manage to do it for one story only.

storyviews = stats[['title', 'views']].sort_values(by=['title','views'])
storyviews = stats[stats["title"] == "Getting Started with TDD"]
storyviews = storyviews[["title","views"]].sort_values(by=['title','views'])
difference = storyviews.set_index('title').diff()
difference = difference.dropna(subset=['views'])
difference

and I got the correct result.

Is there a way to do it in one pass for all the stories?

DATASET

y,m,d,mediumID,title,link,publication,mins,views,reads,readRatio,fans,pubDate,liveDate
2020,06,30,a1777d8bf7e,Swift — Filtering: A Real Example,https://levelup.gitconnected.com/swift-filtering-a-real-example-a1777d8bf7e,Level Up Coding,4 min read,35,13,37.142857142857146,1,2020-06-17,2020-06-26
2020,06,30,6f5fc68b0b43,SwiftUI 2: an overview,https://levelup.gitconnected.com/swiftui-2-an-overview-6f5fc68b0b43,Level Up Coding,3 min read,43,22,51.16279069767442,2,2020-06-24,2020-06-24
2020,07,01,a1777d8bf7e,Swift — Filtering: A Real Example,https://levelup.gitconnected.com/swift-filtering-a-real-example-a1777d8bf7e,Level Up Coding,4 min read,37,13,35.13513513513514,1,2020-06-17,2020-06-26
2020,07,01,6f5fc68b0b43,SwiftUI 2: an overview,https://levelup.gitconnected.com/swiftui-2-an-overview-6f5fc68b0b43,Level Up Coding,3 min read,57,29,50.87719298245614,10,2020-06-24,2020-06-24
2020,07,02,a1777d8bf7e,Swift — Filtering: A Real Example,https://levelup.gitconnected.com/swift-filtering-a-real-example-a1777d8bf7e,Level Up Coding,4 min read,37,13,35.13513513513514,1,2020-06-17,2020-06-26
2020,07,02,6f5fc68b0b43,SwiftUI 2: an overview,https://levelup.gitconnected.com/swiftui-2-an-overview-6f5fc68b0b43,Level Up Coding,3 min read,76,43,56.578947368421055,15,2020-06-24,2020-06-24
2020,07,03,a1777d8bf7e,Swift — Filtering: A Real Example,https://levelup.gitconnected.com/swift-filtering-a-real-example-a1777d8bf7e,Level Up Coding,4 min read,40,13,34.21052631578947,1,2020-06-17,2020-06-26
2020,07,03,6f5fc68b0b43,SwiftUI 2: an overview,https://levelup.gitconnected.com/swiftui-2-an-overview-6f5fc68b0b43,Level Up Coding,3 min read,152,70,46.05263157894737,20,2020-06-24,2020-06-24

Thanks, Nicolas

Answer 1

Could you give this a shot?

cols = ['title', 'views']
storyviews = stats[cols].sort_values(by=cols)
res = storyviews.set_index('title').groupby('title', sort=False).diff().dropna()

Output:

                                   views
title                                   
SwiftUI 2: an overview              14.0
SwiftUI 2: an overview              19.0
SwiftUI 2: an overview              76.0
Swift — Filtering: A Real Example     2.0
Swift — Filtering: A Real Example     0.0
Swift — Filtering: A Real Example     3.0

For plotting the legend, title..., the you might want to ask another question. I don't have an answer. To get you started on the plot, try this.

res.reset_index().groupby('title', sort=False).plot()

Answer 2

So you're just trying to get a sum of all views per title? Here's one-way using dictionary conversion from a pandas DataFrame:

Edited with updated answer in for loop

...
dict = df.groupby('title').groups
for key in dict:
    numpyarray_diff_between = np.diff(np.array(list(dict[key])))
...