Merge pull request #2467 from HHS/ttahub-3451/format_comlog_comdate_h…

…istory

[TTAHUB-3451]format comlog comdate history

src/migrations/20241114000000-format-comlog-comdate-history.js

@@ -0,0 +1,83 @@
+const { prepMigration } = require('../lib/migration');
+
+/** @type {import('sequelize-cli').Migration} */
+module.exports = {
+  async up(queryInterface) {
+    await queryInterface.sequelize.transaction(async (transaction) => {
+      const sessionSig = __filename;
+      await prepMigration(queryInterface, transaction, sessionSig);
+      return queryInterface.sequelize.query(`
+        -- This reformats all historical communicationDate values to mm/dd/yyyy
+        -- 
+        -- Assumptions of preexisting data:
+        -- -always month-day-year
+        -- -always separated by a slash, period, or space [/. ]
+        -- -if there is an extra separator it impacts the year
+        -- -if the third position is at least two characters, it's the year, else it's the fourth position
+        -- -the first four characters of a long year string holds the year
+        --
+        -- These assumptions are based on the data we have, so aren't guaranteed to be correct if this is rerun.
+        -- However the logic is slightly overengineered for extra robustness if new errors show up. At the time of
+        -- writing, this produces all valid strings where to_date(data->>'communicationDate','mm/dd/yyyy') succeeds
+
+        DROP TABLE IF EXISTS comdate_corrections;
+        CREATE TEMP TABLE comdate_corrections
+        AS
+        WITH reseparated AS (
+        SELECT
+          id clid,
+          data->>'communicationDate' orig,
+          -- replace [-. ] seperators (only - has been seen) with / so the subsequent logic always works
+          regexp_replace(data->>'communicationDate','[-. ]','/','g') reseparated
+        FROM "CommunicationLogs"
+        WHERE data->>'communicationDate' !~ '^\\d{2}/\\d{2}/\\d{4}$'
+          AND COALESCE(data->>'communicationDate','') != ''
+        ),
+        date_particles AS (
+        SELECT
+          clid,
+          orig,
+          SPLIT_PART(reseparated,'/',1) month_part,
+          SPLIT_PART(reseparated,'/',2) day_part,
+          -- check where the year part is because sometimes separators between day and year are doubled
+          CASE
+            WHEN LENGTH(SPLIT_PART(reseparated,'/',3)) > 1 THEN SPLIT_PART(reseparated,'/',3)
+            ELSE SPLIT_PART(reseparated,'/',4)
+          END AS year_part
+        FROM reseparated
+        ),
+        padded_particles AS (
+        SELECT
+          clid,
+          orig,
+          LPAD(month_part,2,'0') padded_month,
+          LPAD(day_part,2,'0') padded_day,
+          -- pull out only the leftmost 4 characters, but pad them with the century if we only two chars
+          LPAD(
+            LEFT(year_part,4),
+            4,
+            '20'
+          ) padded_year
+        FROM date_particles
+        )
+        SELECT
+          clid,
+          orig,
+          padded_month || '/' || padded_day || '/' || padded_year reformat
+        FROM padded_particles
+        ;
+
+        UPDATE "CommunicationLogs"
+        SET data = jsonb_set(data, '{communicationDate}', to_jsonb(reformat))
+        FROM comdate_corrections
+        WHERE id = clid
+        ;
+
+      `);
+    });
+  },
+
+  async down() {
+    // no rollbacks
+  },
+};

src/queries/dataRequests/user/communication-logs.sql

@@ -169,14 +169,7 @@ SELECT
     COALESCE(cl.data ->> 'purpose', '') AS "purpose",
     COALESCE(cl.data ->> 'duration', '') AS "duration",
     COALESCE(cl.data ->> 'regionId', '') AS "region",
-    CASE
-		WHEN data ->> 'communicationDate' ~ '^[0-9]{1,2}/[0-9]{1,2}/[0-9]{4}$' THEN TO_DATE(data ->> 'communicationDate', 'MM/DD/YYYY')
-		WHEN data ->> 'communicationDate' ~ '^[0-9]{1,2}/[0-9]{1,2}/[0-9]{2}$' THEN TO_DATE(data ->> 'communicationDate', 'MM/DD/YY')
-		WHEN data ->> 'communicationDate' ~ '^[0-9]{1,2}-[0-9]{1,2}-[0-9]{2}$' THEN TO_DATE(data ->> 'communicationDate', 'MM-DD-YY')
-		WHEN data ->> 'communicationDate' ~ '^[0-9]{1,2}/[0-9]{1,2}//[0-9]{2}$' THEN TO_DATE(regexp_replace(data ->> 'communicationDate', '//', '/'), 'MM/DD/YY')
-		WHEN data ->> 'communicationDate' ~ '^[0-9]{1,2}/[0-9]{1,2}/[0-9]{4}?[0-9]{1,2}.[0-9]{1,2}$' THEN TO_DATE(LEFT(data ->> 'communicationDate', 10), 'MM/DD/YYYY')
-		ELSE NULL
-	END AS "communicationDate",
+    TO_DATE(data ->> 'communicationDate', 'MM/DD/YYYY') "communicationDate",
     COALESCE(cl.data ->> 'pocComplete', '') AS "pocComplete",
     COALESCE(cl.data ->> 'notes', '') AS "notes",
     COALESCE((