Oleg Shpynov

Results 20 comments of Oleg Shpynov

This happened during parallel update - one by crontab schedule, and another process - manual update. We should avoid parallel updates on the system level with some lock file.

See paper Mapping Change in large networks https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0008694&type=printable

@ctrltz is it possible to use `np.log1p` to avoid this problem?

Also tokens_freqs_norm may be zero. What is correct fix for this?

Explain analyse query for the part of query used in bibliographic coupling fetching: ``` explain analyse SELECT ssid_out, ssid_in, crc32id_in FROM sscitations C WHERE (crc32id_out, ssid_out) IN (VALUES (-2004926960, 'eb33b4f5b7ba0f135f1025cac48d7fa26d43668b'),...

Number of citations is treated as negative for some reason, so that log scaling doesn't apply. Data itself: ``` year id ... paper_year count 0 1985 2857475 ... 1985 2...

The issue is related with log scale y axis, when plotting on linear scale everything is okay.