poor search results on related content / related to weighing on claim bid and %20? #108

Closed
opened 2018-08-27 19:54:11 +02:00 by tzarebczan · 14 comments
tzarebczan commented 2018-08-27 19:54:11 +02:00 (Migrated from github.com)

When going to lbry://cz-abl-rhodes#4eff956d91e02c9baae3f1f843a01cadf08b7556, the recommended results are pretty poor (based on the title). Search query: https://lighthouse.lbry.io/search?s=CryptoCandor%20%7C%20Live%20w%20Crypto%20Zombie,%20Sara%20-%20Altcoin%20Buzz%20Ladies,%20&%20Rhode%20Block!&size=20&from=0

I'm guessing this is related to https://github.com/lbryio/lighthouse/issues/32

When going to lbry://cz-abl-rhodes#4eff956d91e02c9baae3f1f843a01cadf08b7556, the recommended results are pretty poor (based on the title). Search query: https://lighthouse.lbry.io/search?s=CryptoCandor%20%7C%20Live%20w%20Crypto%20Zombie,%20Sara%20-%20Altcoin%20Buzz%20Ladies,%20&%20Rhode%20Block!&size=20&from=0 I'm guessing this is related to https://github.com/lbryio/lighthouse/issues/32
tzarebczan commented 2018-08-27 20:07:11 +02:00 (Migrated from github.com)

here's another example on lbry://quick-lunduke-show-update-lunduke-show#4f995883cd68f4561d3b8b5e5744af0571eac384
example

here's another example on lbry://quick-lunduke-show-update-lunduke-show#4f995883cd68f4561d3b8b5e5744af0571eac384 ![example](https://content.screencast.com/users/thomas.zarebczan/folders/Snagit/media/7953ee8c-4334-4f20-9544-22eda566ca73/08.27.2018-14.01.png)
tzarebczan commented 2018-08-27 20:34:34 +02:00 (Migrated from github.com)

Here's an even worse one...lots of the @Porn hits from the community top bids:
image

Here's an even worse one...lots of the @Porn hits from the community top bids: ![image](https://user-images.githubusercontent.com/8120721/44678417-4d683200-aa06-11e8-9803-92291a50ea75.png)
tzarebczan commented 2018-08-27 22:52:21 +02:00 (Migrated from github.com)

Just noticed: https://github.com/lbryio/lbry-desktop/issues/1911

so the query was: https://lighthouse.lbry.io/search?s=GameTales%20 - the %20 (space) is bringing in the porn results. Maybe this is what's causing it to happen on other queries too.

We should try to get this patched up ASAP since we are close to releasing an app with recommended content.

Just noticed: https://github.com/lbryio/lbry-desktop/issues/1911 so the query was: https://lighthouse.lbry.io/search?s=GameTales%20 - the %20 (space) is bringing in the porn results. Maybe this is what's causing it to happen on other queries too. We should try to get this patched up ASAP since we are close to releasing an app with recommended content.
tiger5226 commented 2018-08-27 23:54:10 +02:00 (Migrated from github.com)

@seanyesmunt created another issue for this. He is now filtering out NSFW content appside. #110 check it out. If we can avoid the unintended side effects of effective amount is that preferable? I mean what we have is what we expected by implementing the issue, just not desired.

@seanyesmunt created another issue for this. He is now filtering out NSFW content appside. #110 check it out. If we can avoid the unintended side effects of effective amount is that preferable? I mean what we have is what we expected by implementing the issue, just not desired.
tzarebczan commented 2018-08-28 00:22:04 +02:00 (Migrated from github.com)

These results shouldn't be coming in at all - they are only coming back because of the %20/space in the search. How can we make sure that these characters are ignored? If we are passing search terms, it should only take the terms into account, not any special characters, including spaces.

These results shouldn't be coming in at all - they are only coming back because of the %20/space in the search. How can we make sure that these characters are ignored? If we are passing search terms, it should only take the terms into account, not any special characters, including spaces.
tiger5226 commented 2018-08-28 00:51:27 +02:00 (Migrated from github.com)

there should be no leading or trailing spaces in a search. We can trim both sides of any query that comes in to enforce that.

there should be no leading or trailing spaces in a search. We can trim both sides of any query that comes in to enforce that.
tiger5226 commented 2018-08-28 00:52:34 +02:00 (Migrated from github.com)

I am also worried that we are doing something wrong app side to produce this. @seanyesmunt Any idea how a trailing space is getting passed to the search api?

I am also worried that we are doing something wrong app side to produce this. @seanyesmunt Any idea how a trailing space is getting passed to the search api?
tzarebczan commented 2018-08-28 06:05:01 +02:00 (Migrated from github.com)
It's not only an issue with the trailing space, it's an issue with any spaces. See first example: https://lighthouse.lbry.io/search?s=CryptoCandor%20%7C%20Live%20w%20Crypto%20Zombie,%20Sara%20-%20Altcoin%20Buzz%20Ladies,%20&%20Rhode%20Block!&size=20&from=0
tzarebczan commented 2018-08-28 15:37:38 +02:00 (Migrated from github.com)

If we don't understand how to handle this scenario, we should revert the effective amount changes for the time being.

If we don't understand how to handle this scenario, we should revert the effective amount changes for the time being.
tzarebczan commented 2018-08-29 00:13:08 +02:00 (Migrated from github.com)

Lighthouse IP was reverted to an earlier instance as a workaround.

Lighthouse IP was reverted to an earlier instance as a workaround.
alyssaoc commented 2018-08-29 15:38:47 +02:00 (Migrated from github.com)

Need to understand the root cause before proceeding

Need to understand the root cause before proceeding
tiger5226 commented 2018-09-16 03:56:09 +02:00 (Migrated from github.com)

The root cause was due to the search query not being trimmed. We previously had an issue with internal server errors on specific characters. These characters needed to be escaped. They also needed the trim function called. When I added this the results even with the effective amount worked great.

@alyssaoc This needs coordination with @nikooo777 / @filipnyquist . They did something with the machine and reverted to an old vm? Not sure exactly what happened. The effective amount changes will cause internal server errors if the elasticsearch db is not rebuilt. The IP was changed, so I assume we can merge, pull, build, deploy on the REAL production instance, and then change the IPs back.

@nikooo777 Can you please confirm?

Solved with https://github.com/lbryio/lighthouse/pull/114

The root cause was due to the search query not being trimmed. We previously had an issue with internal server errors on specific characters. These characters needed to be escaped. They also needed the trim function called. When I added this the results even with the effective amount worked great. @alyssaoc This needs coordination with @nikooo777 / @filipnyquist . They did something with the machine and reverted to an old vm? Not sure exactly what happened. The effective amount changes will cause internal server errors if the elasticsearch db is not rebuilt. The IP was changed, so I assume we can merge, pull, build, deploy on the REAL production instance, and then change the IPs back. @nikooo777 Can you please confirm? Solved with https://github.com/lbryio/lighthouse/pull/114
nikooo777 commented 2018-09-17 06:34:07 +02:00 (Migrated from github.com)

Correct!

@filipnyquist is this something you can handle in the following days? I can switch the IP whenever it's ready

Correct! @filipnyquist is this something you can handle in the following days? I can switch the IP whenever it's ready
tiger5226 commented 2018-09-23 16:51:52 +02:00 (Migrated from github.com)

This has been merged. Currently awaiting a deployment to production for testing. Then if all goes well the ip can be switched to the production instance.

This has been merged. Currently awaiting a deployment to production for testing. Then if all goes well the ip can be switched to the production instance.
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: LBRYCommunity/lighthouse.js#108
No description provided.