Merge pull request #91 from CartoDB/upgrade-version-0-7-4

Update Makefile and NEWS.md for new version

Makefile

@@ -1,7 +1,7 @@
 # cartodb/Makefile
 
 EXTENSION = cartodb
-EXTVERSION = 0.7.2
+EXTVERSION = 0.7.4
 
 SED = sed
 
@@ -36,6 +36,8 @@ UPGRADABLE = \
   0.6.0 \
   0.7.0 \
   0.7.1 \
+  0.7.2 \
+  0.7.3 \
   $(EXTVERSION)dev \
   $(EXTVERSION)next \
   $(END)

NEWS.md

@@ -1,3 +1,17 @@
+0.7.4 (2015-06-29)
+------------------
+* Adds new function CDB_QueryTablesText that can deal with "schema.table_name"
+  longer than 63 chars.
+* Adds a set of statistical functions:
+  - CDB_DistType
+  - CDB_DistinctMeasure
+  - CDB_EqualIntervalBins
+
+0.7.3 (2015-03-03)
+------------------
+* Fix upgrade of CDB_StringToDate function
+* Add a test for to validate CDB_TableMetadataTouch usage with OID
+
 0.7.2 (2015-03-03)
 ------------------
 * Fix conversion of strings to datetime

scripts-available/CDB_DistType.sql

@@ -0,0 +1,122 @@
+--
+-- CDB_DistType classifies the histograms of a column into
+-- one of the basic types listed by Galtung: http://druedin.com/2012/12/08/galtungs-ajus-system/
+-- 
+-- Future improvements:
+--    variable number of bins (7 is baked in right now)
+--    catch the number of items to ensure that the sample is large enough
+--
+-- Refs:
+--    1. width_bucket/histograms: http://tapoueh.org/blog/2014/02/21-PostgreSQL-histogram
+--    2. R implementation: https://github.com/cran/agrmt
+
+CREATE OR REPLACE FUNCTION CDB_DistType ( in_array NUMERIC[] ) RETURNS text as $$
+DECLARE
+    element_count INT4;
+    minv numeric;
+    maxv numeric;
+    bins numeric[];
+    freqs numeric[];
+    ajus INT[];
+    freq INT4;
+    signature text;
+    i INT := 1;
+BEGIN
+    SELECT min(e), max(e), count(e) INTO minv, maxv, element_count FROM ( SELECT unnest(in_array) e ) x;
+
+    IF abs(maxv - minv) < 1e-7 THEN -- if max and min are nearly equal, call if 'F' (make relative to maxv?)
+        signature = 'F';
+    ELSE
+        -- Calculate bins and count in bins
+        EXECUTE 'WITH stats as (
+            SELECT min(e) as minv,
+                   max(e) as maxv,
+                   count(e) as total
+            FROM (SELECT unnest($1) e) x
+            WHERE e is not null
+        ),
+        hist as (
+            SELECT width_bucket(e, s.minv, s.maxv, 7) bucket,
+                   count(*) freq
+            FROM (SELECT unnest($1) e) x, stats s
+            WHERE e is not null
+            GROUP BY 1
+            ORDER BY 1
+        )
+        SELECT array_agg(round(100.0 * hist.freq::numeric / stats.total::numeric,1)) freqs,
+               array_agg(hist.bucket) buckets
+        FROM hist, stats'
+        INTO freqs, bins
+        USING in_array;
+
+        LOOP
+            IF i < 7 THEN
+                ajus[i] = CASE WHEN freqs[i] > freqs[i+1] THEN -1
+                               WHEN abs(freqs[i] - freqs[i+1]) <= 0.05 THEN 0
+                               ELSE 1 END;
+            ELSE
+                EXIT;
+            END IF;
+            i := i + 1;
+        END LOOP;
+
+        signature = _CDB_DistTypeClassify(ajus);
+    END IF;
+
+    RETURN signature;
+END;
+$$ language plpgsql IMMUTABLE;
+
+-- Classify data into AJUSFL
+
+CREATE OR REPLACE FUNCTION _CDB_DistTypeClassify ( in_array INT[] ) RETURNS text as $$
+DECLARE
+    element_count INT4;
+    maxv numeric;
+    minv numeric;
+    uniques INT[];
+    type text;
+BEGIN
+    SELECT max(e), min(e) INTO maxv, minv FROM ( SELECT unnest(in_array) e ) x;
+
+    IF (maxv = 0 AND minv = 0) THEN
+        type = 'F';
+    ELSIF maxv < 1 THEN
+        type = 'L';
+    ELSIF minv > -1 THEN
+        type = 'J';
+    ELSE
+        -- Get distinct elements ordered by original position
+        EXECUTE 'WITH b AS (
+            SELECT a
+            FROM (SELECT unnest($1) a) x
+        ),
+        c AS (
+            SELECT a, row_number() OVER () r
+            FROM b
+        ),
+        d AS (
+            SELECT DISTINCT a
+            FROM c
+        ),
+        e AS (
+            SELECT a FROM d ORDER BY (
+                SELECT r FROM c WHERE d.a = c.a ORDER BY r ASC LIMIT 1
+            ) ASC)
+        SELECT array_agg(a) FROM e'
+        INTO uniques
+        USING in_array;
+
+        -- Decide if it's an A, U, or other
+        IF (uniques = ARRAY[1,-1] OR uniques = ARRAY[1,0,-1] OR uniques = ARRAY[1,-1,0] OR uniques = ARRAY[0,1,-1]) THEN
+            type = 'A';
+        ELSIF (uniques = ARRAY[-1,1] OR uniques = ARRAY[-1,0,1] OR uniques = ARRAY[-1,1,0] OR uniques = ARRAY[0,-1,1]) THEN
+            type = 'U';
+        ELSE
+            type = 'S';
+        END IF;
+    END IF;
+
+    RETURN type;
+END;
+$$ language plpgsql IMMUTABLE;

scripts-available/CDB_DistinctMeasure.sql

@@ -0,0 +1,46 @@
+--
+-- CDB_DistinctMeasure 
+--     calculates the fraction of rows in the 10 most common distinct categories
+--     returns true if the number of rows in these 10 categories is >= 0.9 * total number of rows
+-- 
+-- 
+
+CREATE OR REPLACE FUNCTION CDB_DistinctMeasure ( in_array text[], threshold numeric DEFAULT null ) RETURNS numeric as $$
+DECLARE
+    element_count INT4;
+    maxval numeric;
+    passes numeric;
+BEGIN
+    SELECT count(e) INTO element_count FROM ( SELECT unnest(in_array) e ) x;
+
+    -- count number of occurrences per bin
+    -- calculate the normalized cumulative sum
+    -- return the max value: which corresponds nth entry 
+    -- for n <= 10 depending on # of distinct values
+    EXECUTE 'WITH a As (
+              SELECT
+                count(*) cnt
+              FROM
+                (SELECT * FROM unnest($2) e ) x
+              WHERE e is not null
+              GROUP BY e
+              ORDER BY cnt DESC
+            ),
+            b As (
+              SELECT
+                sum(cnt) OVER (ORDER BY cnt DESC) / $1 As cumsum
+              FROM a
+              LIMIT 10
+            )
+            SELECT max(cumsum) maxval FROM b'
+            INTO maxval
+            USING element_count, in_array;
+    IF threshold is null THEN
+        passes = maxval;
+    ELSE
+        passes = CASE WHEN (maxval >= threshold) THEN 1 ELSE 0 END;
+    END IF;
+
+    RETURN passes;
+END;
+$$ language plpgsql IMMUTABLE;

scripts-available/CDB_EqualIntervalBins.sql

@@ -0,0 +1,37 @@
+--
+-- Calculate the equal interval bins for a given column
+--
+-- @param in_array A numeric array of numbers to determine the best
+--                   to determine the bin boundary
+--
+-- @param breaks The number of bins you want to find.
+--  
+--
+-- Returns: upper edges of bins
+-- 
+--
+
+CREATE OR REPLACE FUNCTION CDB_EqualIntervalBins ( in_array NUMERIC[], breaks INT ) RETURNS NUMERIC[] as $$
+DECLARE 
+    diff numeric;
+    min_val numeric;
+    max_val numeric;
+    tmp_val numeric;
+    i INT := 1;
+    reply numeric[];
+BEGIN
+    SELECT min(e), max(e) INTO min_val, max_val FROM ( SELECT unnest(in_array) e ) x WHERE e IS NOT NULL;
+    diff = (max_val - min_val) / breaks::numeric;
+    LOOP
+        IF i < breaks THEN
+            tmp_val = min_val + i::numeric * diff;
+            reply = array_append(reply, tmp_val);
+            i := i+1;
+        ELSE
+            reply = array_append(reply, max_val);
+            EXIT;
+        END IF;
+    END LOOP;
+    RETURN reply;
+END;
+$$ language plpgsql IMMUTABLE;

scripts-available/CDB_QuantileBins.sql

@@ -15,18 +15,29 @@ DECLARE
     i INT := 1; 
     reply numeric[]; 
 BEGIN 
-    -- get our unique values
-    SELECT array_agg(e) INTO in_array FROM (SELECT unnest(in_array) e GROUP BY e ORDER BY e ASC) x;
-    -- get the total size of our row
-    element_count := array_upper(in_array, 1) - array_lower(in_array, 1); 
+    -- sort our values
+    SELECT array_agg(e) INTO in_array FROM (SELECT unnest(in_array) e ORDER BY e ASC) x;
+    -- get the total size of our data
+    element_count := array_length(in_array, 1); 
     break_size :=  element_count::numeric / breaks;
     -- slice our bread
     LOOP  
-        IF i > breaks THEN  EXIT;  END IF;  
-        SELECT e INTO tmp_val FROM ( SELECT unnest(in_array) e LIMIT 1 OFFSET round(break_size * i)) x;
+        IF i < breaks THEN
+            IF break_size * i % 1 > 0 THEN
+                SELECT e INTO tmp_val FROM ( SELECT unnest(in_array) e LIMIT 1 OFFSET ceil(break_size * i) - 1) x;
+            ELSE
+                SELECT avg(e) INTO tmp_val FROM ( SELECT unnest(in_array) e LIMIT 2 OFFSET ceil(break_size * i) - 1 ) x;
+            END IF;
+        ELSIF i = breaks THEN
+            -- select the last value
+            SELECT max(e) INTO tmp_val FROM ( SELECT unnest(in_array) e ) x;
+        ELSE
+            EXIT;
+        END IF;
+
         reply = array_append(reply, tmp_val);
-        i := i+1; 
-    END LOOP; 
-    RETURN reply; 
+        i := i+1;
+    END LOOP;
+    RETURN reply;
 END; 
-$$ language plpgsql IMMUTABLE;
+$$ language plpgsql IMMUTABLE;

scripts-available/CDB_QueryTables.sql

@@ -2,12 +2,12 @@
 --
 -- Requires PostgreSQL 9.x+
 --
-CREATE OR REPLACE FUNCTION CDB_QueryTables(query text)
-RETURNS name[]
+CREATE OR REPLACE FUNCTION CDB_QueryTablesText(query text)
+RETURNS text[]
 AS $$
 DECLARE
   exp XML;
-  tables NAME[];
+  tables text[];
   rec RECORD;
   rec2 RECORD;
 BEGIN
@@ -41,11 +41,11 @@ BEGIN
           xpath('//x:Relation-Name/text()', exp, ARRAY[ARRAY['x', 'http://www.postgresql.org/2009/explain']]) as x,
           xpath('//x:Relation-Name/../x:Schema/text()', exp, ARRAY[ARRAY['x', 'http://www.postgresql.org/2009/explain']]) as s
       )
-      SELECT unnest(x)::name as p, unnest(s)::name as sc from inp
+      SELECT unnest(x) as p, unnest(s) as sc from inp
     LOOP
       -- RAISE DEBUG 'tab: %', rec2.p;
       -- RAISE DEBUG 'sc: %', rec2.sc;
-      tables := array_append(tables, (rec2.sc || '.' || rec2.p)::name);
+      tables := array_append(tables, (rec2.sc || '.' || rec2.p));
     END LOOP;
 
     -- RAISE DEBUG 'Tables: %', tables;
@@ -65,3 +65,14 @@ BEGIN
   return tables;
 END
 $$ LANGUAGE 'plpgsql' VOLATILE STRICT;
+
+
+-- Keep CDB_QueryTables with same signature for backwards compatibility.
+-- It should probably be removed in the future.
+CREATE OR REPLACE FUNCTION CDB_QueryTables(query text)
+RETURNS name[]
+AS $$
+BEGIN
+  RETURN CDB_QueryTablesText(query)::name[];
+END
+$$ LANGUAGE 'plpgsql' VOLATILE STRICT;

scripts-available/CDB_StringToDate.sql

@@ -1,5 +1,6 @@
 -- Convert string to date
 --
+DROP FUNCTION IF EXISTS CDB_StringToDate(character varying);
 CREATE OR REPLACE FUNCTION CDB_StringToDate(input character varying)
 RETURNS TIMESTAMP AS $$
 DECLARE output TIMESTAMP;

scripts-enabled/240-CDB_EqualIntervalBins.sql

@@ -0,0 +1 @@
+../scripts-available/CDB_EqualIntervalBins.sql

scripts-enabled/CDB_DistType.sql

@@ -0,0 +1 @@
+../scripts-available/CDB_DistType.sql

scripts-enabled/CDB_DistinctMeasure.sql

@@ -0,0 +1 @@
+../scripts-available/CDB_DistinctMeasure.sql

test/CDB_DistTypeTest.sql

@@ -0,0 +1,4 @@
+WITH data AS (
+    SELECT pow(x,3)::numeric x FROM generate_series(-100,100) x
+    ) 
+SELECT CDB_DistType(array_agg(x)) FROM data

test/CDB_DistTypeTest_expect

@@ -0,0 +1 @@
+A

test/CDB_DistinctMeasureTest.sql

@@ -0,0 +1,20 @@
+-- a - j add up to 89%, k-m add up to 11%
+WITH a As (
+	SELECT (
+		repeat('a',12) || 
+		repeat('b',11) || 
+		repeat('c',11) || 
+		repeat('d',10) || 
+		repeat('e',10) || 
+		repeat('f',9) || 
+		repeat('g',8) || 
+		repeat('h',7) || 
+		repeat('i',6) || 
+		repeat('j',5) || 
+		repeat('k',4) || 
+		repeat('l',4) || 
+		repeat('m',3)
+		)::text AS x 
+	) 
+
+SELECT CDB_DistinctMeasure(string_to_array(x,null),0.90) from a

test/CDB_DistinctMeasureTest_expect

@@ -0,0 +1 @@
+0

test/CDB_EqualIntervalBinsTest.sql

@@ -0,0 +1,5 @@
+WITH data AS (
+    SELECT array_agg(x::numeric) s FROM generate_series(1,300) x 
+        WHERE x % 5 != 0 AND x % 7 != 0
+    ) 
+SELECT round(unnest(CDB_EqualIntervalBins(s, 7)),7) FROM data

test/CDB_EqualIntervalBinsTest_expect

@@ -0,0 +1,7 @@
+43.5714286
+86.1428571
+128.7142857
+171.2857143
+213.8571429
+256.4285714
+299.0000000

test/CDB_QuantileBinsTest_expect

@@ -1,7 +1,7 @@
-16
+13
 29
 43
 57
 71
-83
+86
 99

test/CDB_QueryTablesTest_expect

@@ -4,10 +4,14 @@ CREATE table "my'tab;le" as select 1|{}
 SELECT a.oid, b.oid FROM pg_class a, pg_class b|{pg_catalog.pg_class}
 SELECT 1 as col1; select 2 as col2|{}
 WARNING:  CDB_QueryTables cannot explain query: select 1 from nonexistant (42P01: relation "nonexistant" does not exist)
+CONTEXT:  PL/pgSQL function cdb_querytables(text) line 3 at RETURN
 ERROR:  relation "nonexistant" does not exist
+CONTEXT:  PL/pgSQL function cdb_querytables(text) line 3 at RETURN
 begin; select * from pg_class; commit;|{pg_catalog.pg_class}
 WARNING:  CDB_QueryTables cannot explain query: select * from test (42P01: relation "test" does not exist)
+CONTEXT:  PL/pgSQL function cdb_querytables(text) line 3 at RETURN
 ERROR:  relation "test" does not exist
+CONTEXT:  PL/pgSQL function cdb_querytables(text) line 3 at RETURN
 WITH a AS (select * from pg_class) select * from a|{pg_catalog.pg_class}
 CREATE SCHEMA
 CREATE TABLE

test/extension/test.sh

@@ -302,6 +302,21 @@ function test_cdb_tablemetadatatouch() {
     sql "SELECT CDB_TableMetadataTouch('\"public\".touch_example');"
     sql "SELECT CDB_TableMetadataTouch('\"public\".\"touch_example\"');"
 
+    # Works with OID
+    sql postgres "SELECT tabname from CDB_TableMetadata;" should 'touch_example'
+    sql postgres "SELECT count(*) from CDB_TableMetadata;" should 1
+    TABLE_OID=`${CMD} -U postgres ${DATABASE} -c "SELECT attrelid FROM pg_attribute WHERE attrelid = 'touch_example'::regclass limit 1;" -A -t`
+
+    # quoted OID works
+    sql "SELECT CDB_TableMetadataTouch('${TABLE_OID}');"
+    sql postgres "SELECT tabname from CDB_TableMetadata;" should 'touch_example'
+    sql postgres "SELECT count(*) from CDB_TableMetadata;" should 1
+
+    # non quoted OID works
+    sql "SELECT CDB_TableMetadataTouch(${TABLE_OID});"
+    sql postgres "SELECT tabname from CDB_TableMetadata;" should 'touch_example'
+    sql postgres "SELECT count(*) from CDB_TableMetadata;" should 1
+
     #### test tear down
     sql 'DROP TABLE touch_example;'
 }