统计你的关注者

A recent wave of users have questioned the validity of the displayed number of followers on their profiles. At one point, their follower number increased, some even by the thousands, but then a few weeks later went back down to around where the user originally remembered. Some users were concerned: Were their profiles hacked? Did a wave of users suddenly follow and then unfollow them? The answer is no—it was a data problem.

最近一波用户对其个人资料中显示的关注者数量的有效性提出了质疑。有一段时间，他们的关注者数量增加了，甚至有些增加了数千个，但几周后又回到了用户最初记得的水平。一些用户感到担忧：他们的个人资料被黑客入侵了吗？是否有一波用户突然关注然后取消关注他们？答案是否定的，这是一个数据问题。

When Medium started to shift toward more relational content, we noticed that some users were missing authors in their “Latest From Following” section, which should show the followed collections and authors that had recently published. Since this wasn’t true for all users, we had to look into the data that drives this section of Medium.

当Medium开始转向更多的关系内容时，我们注意到一些用户在他们的“最新关注”部分中缺少作者，该部分应该显示最近发布的关注的集合和作者。由于这并不适用于所有用户，我们不得不查看驱动Medium此部分的数据。

Upon investigation, we noticed a discrepancy in a key column that the “Latest From Following” section uses that the rest of Medium does not. In turn, a backfill had to be done so the key column would properly reflect a user’s followings. After the backfill finished, users started reporting an uptick in followers and the number of people they have followed, even though they had not followed any new users. That’s when we noticed that the original backfill had done its intended purpose while also breaking it.

经过调查，我们注意到“最新关注”部分使用了一个关键列，而Medium的其他部分没有使用。因此，必须进行回填，以便关键列能正确反映用户的关注。回填完成后，用户开始报告关注者增加以及他们关注的人数增加，尽管他们没有关注任何新用户。这时我们注意到原始回填已经完成了其预期的目的，同时也破坏了它。

A backfill is supposed to be the process in correcting data or making the data whole without modifying its existing core components. The first backfill did fix the key column needed for the table that “Latest From Following” depended on, but it also ended up modifying another table that contained the displayed number o...