General improvement
Jun. 6th, 2019 07:28 am![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
The weather has been clearing up somewhat although, as is typical for Britain, it remains changeable. My right side continues to heal from whatever happened to it. I am still favoring my left arm even more than usual but I no longer face some movements with trepidation and I expect that after another week or so the issue will barely be perceptible.
At work I seem to have finally figured a fix for the mysteriously failing test. This is most agreeable as it threatened to block release and I had little idea what was happening. Indeed, the fix is just the latest in a series of optimistic guesses about some dimension of the issue.
With that failing test search results were sometimes not being found by Apache Lucene. Hibernate Search's TRACE-level logging recorded writing values to the fields appropriately but some tokenized entries sometimes did not get the corresponding search terms written to the index. I realized that the issue had been observed only for fields that have both tokenized and untokenized values written so I adjusted our code generation to not produce model objects that do such mixing. For the meantime we can live with that change and this appears to have fixed the test on all the servers we have tried so far. Whether the problem arising from writing such a mix of values points to a bug in our code, Hibernate Search or Apache Lucene I've no idea. It used to work fine on earlier versions.
Yesterday evening's workout went fairly well. I have a hesitant theory that I tend to have an easier workout after a more successful workday.
At work I seem to have finally figured a fix for the mysteriously failing test. This is most agreeable as it threatened to block release and I had little idea what was happening. Indeed, the fix is just the latest in a series of optimistic guesses about some dimension of the issue.
With that failing test search results were sometimes not being found by Apache Lucene. Hibernate Search's TRACE-level logging recorded writing values to the fields appropriately but some tokenized entries sometimes did not get the corresponding search terms written to the index. I realized that the issue had been observed only for fields that have both tokenized and untokenized values written so I adjusted our code generation to not produce model objects that do such mixing. For the meantime we can live with that change and this appears to have fixed the test on all the servers we have tried so far. Whether the problem arising from writing such a mix of values points to a bug in our code, Hibernate Search or Apache Lucene I've no idea. It used to work fine on earlier versions.
Yesterday evening's workout went fairly well. I have a hesitant theory that I tend to have an easier workout after a more successful workday.
no subject
Date: 2019-06-06 11:04 am (UTC)