Remove duplicates from fluview_clinical table

src/ddl/fluview.sql

@@ -269,8 +269,8 @@ CREATE TABLE `fluview_clinical` (
   `percent_a` double DEFAULT NULL,
   `percent_b` double DEFAULT NULL,
   PRIMARY KEY (`id`),
+  UNIQUE KEY `issue` (`issue`, `epiweek`, `region`),
   KEY `release_date` (`release_date`),
-  KEY `issue` (`issue`),
   KEY `epiweek` (`epiweek`),
   KEY `region` (`region`),
   KEY `lag` (`lag`)

src/ddl/migrations/fluview_clinical_v0.1.sql

@@ -0,0 +1,115 @@
+USE epidata;
+
+-- Create new `fluview_clinical` table with proper unique constraint.
+CREATE TABLE `fluview_clinical_v2` (
+    `id` int(11) NOT NULL AUTO_INCREMENT,
+    `release_date` date NOT NULL,
+    `issue` int(11) NOT NULL,
+    `epiweek` int(11) NOT NULL,
+    `region` varchar(12) NOT NULL,
+    `lag` int(11) NOT NULL,
+    `total_specimens` int(11) NOT NULL,
+    `total_a` int(11) DEFAULT NULL,
+    `total_b` int(11) DEFAULT NULL,
+    `percent_positive` double DEFAULT NULL,
+    `percent_a` double DEFAULT NULL,
+    `percent_b` double DEFAULT NULL,
+    PRIMARY KEY (`id`),
+    UNIQUE KEY `issue` (`issue`, `epiweek`, `region`),
+    KEY `release_date` (`release_date`),
+    KEY `epiweek` (`epiweek`),
+    KEY `region` (`region`),
+    KEY `lag` (`lag`)
+) ENGINE = InnoDB DEFAULT CHARSET = utf8;
+
+
+-- Insert unique rows from `fluview_clinical` into `fluview_clinical_v2`.
+-- This is done in order to reset ID counter and fill gaps betwen row's ids.
+INSERT INTO
+    fluview_clinical_v2(
+        `release_date`,
+        `issue`,
+        `epiweek`,
+        `region`,
+        `lag`,
+        `total_specimens`,
+        `total_a`,
+        `total_b`,
+        `percent_positive`,
+        `percent_a`,
+        `percent_b`
+    )
+SELECT
+    min_release_date release_date,
+    tmp.issue,
+    tmp.epiweek,
+    tmp.region,
+    tmp.lag,
+    tmp.total_specimens,
+    tmp.total_a,
+    tmp.total_b,
+    tmp.percent_positive,
+    tmp.percent_a,
+    tmp.percent_b
+FROM
+    (
+        -- get data associated with the most recent `release_date` for each unique `(epiweek, issue, region)` key
+        SELECT
+            s.release_date,
+            s.issue,
+            s.epiweek,
+            s.region,
+            s.lag,
+            s.total_specimens,
+            s.total_a,
+            s.total_b,
+            s.percent_positive,
+            s.percent_a,
+            s.percent_b
+        FROM
+            (
+                SELECT
+                    fc.release_date,
+                    fc.issue,
+                    fc.epiweek,
+                    fc.region,
+                    fc.lag,
+                    fc.total_specimens,
+                    fc.total_a,
+                    fc.total_b,
+                    fc.percent_positive,
+                    fc.percent_a,
+                    fc.percent_b,
+                    ROW_NUMBER() OVER(
+                        PARTITION BY fc.epiweek,
+                        fc.issue,
+                        fc.region
+                        ORDER BY
+                            fc.release_date DESC
+                    ) as row_num
+                FROM
+                    fluview_clinical fc
+            ) s
+        WHERE
+            s.row_num = 1
+    ) tmp
+    JOIN (
+        -- JOIN to recover first/least `release_date` because thats what the acquisition process does: https://github.com/cmu-delphi/delphi-epidata/blob/7fd20cd5c34b33c2310be67867b46a91aa840be9/src/acquisition/fluview/fluview_update.py#L326
+        SELECT
+            MIN(fc.release_date) as min_release_date,
+            fc.issue,
+            fc.epiweek,
+            fc.region
+        FROM
+            fluview_clinical fc
+        GROUP BY
+            fc.issue,
+            fc.epiweek,
+            fc.region
+    ) rel_date ON tmp.issue = rel_date.issue
+    AND tmp.epiweek = rel_date.epiweek
+    AND tmp.region = rel_date.region;
+
+DROP TABLE fluview_clinical;
+
+ALTER TABLE fluview_clinical_v2 RENAME fluview_clinical;