html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,issue,performed_via_github_app
https://github.com/simonw/datasette/issues/190#issuecomment-377457214,https://api.github.com/repos/simonw/datasette/issues/190,377457214,MDEyOklzc3VlQ29tbWVudDM3NzQ1NzIxNA==,9599,2018-03-30T06:31:15Z,2018-03-30T06:31:15Z,OWNER,"Fixed! https://datasette-issue-190-compound-pks-second-fix.now.sh/compound-pks-8e99805/compound_three_primary_keys?_next=b%2Cx%2Cd&content__contains=d now correctly shows `b,y,d` as the first row on the page.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",309558826,
https://github.com/simonw/datasette/issues/190#issuecomment-377457087,https://api.github.com/repos/simonw/datasette/issues/190,377457087,MDEyOklzc3VlQ29tbWVudDM3NzQ1NzA4Nw==,9599,2018-03-30T06:30:23Z,2018-03-30T06:30:23Z,OWNER,"Interestingly, in deploying a copy of the database to demonstrate this final bug fix I had to use the `--force` argument like so:

    datasette publish now --branch=master compound-pks.db --force

This is because `now` had already deployed a Dockerfile referencing `--branch=master` once already, so it thought nothing had changed and it could re-use that last deployment.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",309558826,
https://github.com/simonw/datasette/issues/190#issuecomment-377454591,https://api.github.com/repos/simonw/datasette/issues/190,377454591,MDEyOklzc3VlQ29tbWVudDM3NzQ1NDU5MQ==,9599,2018-03-30T06:11:59Z,2018-03-30T06:11:59Z,OWNER,"Re-opening this issue: my fix doesn't play nicely with extra filter arguments.

Consider this page: https://datasette-issue-190-compound-pks-not-quite-fixed.now.sh/compound-pks-8e99805/compound_three_primary_keys?content__contains=d

The next link is to `?_next=f%2Cz%2Ct&content__contains=z` (that's next of `f,z,t`) but that gives us https://datasette-issue-190-compound-pks-not-quite-fixed.now.sh/compound-pks-8e99805/compound_three_primary_keys?_next=b%2Cx%2Cd&content__contains=d which shows `a,a,d` at the top.

Sure enough, the generated SQL looks like this: https://datasette-issue-190-compound-pks-not-quite-fixed.now.sh/compound-pks-8e99805?sql=select+%2A+from+compound_three_primary_keys+where+%22content%22+like+%3Ap0+and+%28%5Bpk1%5D+%3E+%3Ap0%29%0A++or%0A%28%5Bpk1%5D+%3D+%3Ap0+and+%5Bpk2%5D+%3E+%3Ap1%29%0A++or%0A%28%5Bpk1%5D+%3D+%3Ap0+and+%5Bpk2%5D+%3D+%3Ap1+and+%5Bpk3%5D+%3E+%3Ap2%29+order+by+pk1%2C+pk2%2C+pk3+limit+101&p0=%25d%25&p1=b&p2=x&p3=d

	select * from compound_three_primary_keys where ""content"" like :p0 and ([pk1] > :p0)
	  or
	([pk1] = :p0 and [pk2] > :p1)
	  or
	([pk1] = :p0 and [pk2] = :p1 and [pk3] > :p2) order by pk1, pk2, pk3 limit 101

The parameters here are confused. The :p0 should be reserved just for the like clause - the other parameters should be p1, p2 and p3 (not p0, p1 and p2).","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",309558826,