#+TITLE: Bugs and tasks

* TODO Construct SPARQL query results from JSON instead of CSV   :minor:easy:
  <2021-12-24 Fri>
  sparql-query-results in visualize-schema.scm should retrieve SPARQL
  query results in JSON instead of CSV. CSV ignores type information
  and returns everything as a string. This results in unnecessary
  conversions such as using string->number.
* TODO Specification of field alias in define-dump is a leaky abstraction :wish:
  <2021-12-24 Fri>
  In define-dump expressions, database fields can be referenced by
  (field TABLE COLUMN ALIAS) in addition to (field TABLE
  COLUMN). Specification of the field alias in the first form is a
  leak in the abstraction. There is no need for the caller to know
  anything about field aliases. define-dump should be able to set up
  aliases transparently.
* TODO Guess string encodings                                         :grave:
  <2021-12-24 Fri>
  Over the years, strings have been stored in the database using
  different character encodings. That is why when they are all
  interpreted in UTF-8, they appear to have unprintable
  characters. Somehow, guess the encoding of the retrieved strings and
  compensate.
* TODO Fix table size estimates                                  :minor:easy:
  <2021-12-24 Fri>
  The tables function in dump.scm estimates the table size from the
  data_length column of the information_schema.tables table. But, this
  is correct only for MyISAM tables. Generalize it.
* TODO Dump table and field annotations to RDF                   :minor:easy:
  <2021-12-27 Mon>
  The TableComments and TableFieldAnnotations tables partially
  document the database schema. GeneNetwork 1 uses this to generate
  [[http://gn1.genenetwork.org/webqtl/main.py?FormID=schemaShowPage]]. We
  dump this to an org mode format in the dump-schema-annotations
  function. Dump these tables to RDF instead so that they can be shown
  as tooltip text in our database visualization.