Añadir kit_eda

2025-08-23 01:09:18 -06:00 · 2025-08-23 01:09:18 -06:00 · dd433d5ab2
commit dd433d5ab2
parent ad396d79c8
1 changed files with 140 additions and 0 deletions
--- a/140
+++ b/140
@ -0,0 +1,140 @@
+/* ==========================================================
+   KIT EDA POSTGRESQL – 15 QUERIES CLAVE
+   Autor: NeoAnd
+
+   📌 Instrucciones rápidas:
+
+-- Ejecuta:
+SELECT run_eda_auto_export('public','clientes');
+
+-- Luego en DBeaver:
+-- Busca en el navegador las tablas temporales que empiezan con "tmp_"
+-- Ej: tmp_total_rows, tmp_nulls, tmp_top10_micolumna, tmp_stats_micolumna, tmp_time_midate, tmp_corr
+
+Ejecuta por bloque (Cmd+Enter en Mac, Ctrl+Enter en Windows).
+   ========================================================== */
+
+
+/* ==========================================================
+   Función EDA Autoadaptativa con Export
+   Autor: Anderson + Copilot
+   Uso:   SELECT run_eda_auto_export('schema','tabla');
+   ========================================================== */
+
+CREATE OR REPLACE FUNCTION run_eda_auto_export(p_schema TEXT, p_table TEXT)
+RETURNS void AS
+
+
+DECLARE
+    r RECORD;
+    col_num TEXT[];
+    col_cat TEXT[];
+    col_date TEXT[];
+    sql_cmd TEXT;
+BEGIN
+    -- 🔍 Clasificación automática de columnas
+    SELECT array_agg(column_name) FILTER (WHERE data_type LIKE '%int%' OR data_type LIKE '%numeric%' OR data_type LIKE '%double%')
+    INTO col_num
+    FROM information_schema.columns
+    WHERE table_schema = p_schema AND table_name = p_table;
+
+    SELECT array_agg(column_name) FILTER (WHERE data_type IN ('character varying','text','uuid'))
+    INTO col_cat
+    FROM information_schema.columns
+    WHERE table_schema = p_schema AND table_name = p_table;
+
+    SELECT array_agg(column_name) FILTER (WHERE data_type LIKE '%date%' OR data_type LIKE '%timestamp%')
+    INTO col_date
+    FROM information_schema.columns
+    WHERE table_schema = p_schema AND table_name = p_table;
+
+    -- 📊 1. Conteo total
+    EXECUTE format('DROP TABLE IF EXISTS tmp_total_rows;
+                    CREATE TEMP TABLE tmp_total_rows AS
+                    SELECT COUNT(*) AS total FROM %I.%I',
+                   p_schema, p_table);
+
+    -- 📊 2. Nulos por columna
+    sql_cmd := '';
+    FOR r IN
+        SELECT column_name
+        FROM information_schema.columns
+        WHERE table_schema = p_schema AND table_name = p_table
+    LOOP
+        sql_cmd := sql_cmd || format(
+            'SELECT %L AS columna, COUNT(*) AS nulls FROM %I.%I WHERE %I IS NULL UNION ALL ',
+            r.column_name, p_schema, p_table, r.column_name
+        );
+    END LOOP;
+    sql_cmd := left(sql_cmd, length(sql_cmd)-11); -- quitar último UNION ALL
+    EXECUTE format('DROP TABLE IF EXISTS tmp_nulls;
+                    CREATE TEMP TABLE tmp_nulls AS %s', sql_cmd);
+
+    -- 📊 3. Top 10 valores más frecuentes por columna categórica
+    IF col_cat IS NOT NULL THEN
+        FOREACH r IN ARRAY col_cat LOOP
+            EXECUTE format(
+              'DROP TABLE IF EXISTS tmp_top10_%I;
+               CREATE TEMP TABLE tmp_top10_%I AS
+               SELECT %I, COUNT(*) AS freq
+               FROM %I.%I
+               GROUP BY %I
+               ORDER BY freq DESC
+               LIMIT 10',
+               r, r, r, p_schema, p_table, r
+            );
+        END LOOP;
+    END IF;
+
+    -- 📊 4. Estadísticas descriptivas de columnas numéricas
+    IF col_num IS NOT NULL THEN
+        FOREACH r IN ARRAY col_num LOOP
+            EXECUTE format(
+              'DROP TABLE IF EXISTS tmp_stats_%I;
+               CREATE TEMP TABLE tmp_stats_%I AS
+               SELECT MIN(%I) AS minimo,
+                      MAX(%I) AS maximo,
+                      AVG(%I) AS media,
+                      PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY %I) AS mediana
+               FROM %I.%I',
+               r, r, r, r, r, r, r, p_schema, p_table
+            );
+        END LOOP;
+    END IF;
+
+    -- 📊 5. Distribución temporal por columnas de fecha
+    IF col_date IS NOT NULL THEN
+        FOREACH r IN ARRAY col_date LOOP
+            EXECUTE format(
+              'DROP TABLE IF EXISTS tmp_time_%I;
+               CREATE TEMP TABLE tmp_time_%I AS
+               SELECT date_trunc(''day'', %I) AS dia, COUNT(*)
+               FROM %I.%I
+               GROUP BY dia
+               ORDER BY dia',
+               r, r, r, p_schema, p_table
+            );
+        END LOOP;
+    END IF;
+
+    -- 📊 6. Correlaciones numéricas
+    IF array_length(col_num,1) >= 2 THEN
+        EXECUTE 'DROP TABLE IF EXISTS tmp_corr;
+                 CREATE TEMP TABLE tmp_corr (col_x TEXT, col_y TEXT, correlacion NUMERIC)';
+        FOR i IN 1..array_length(col_num,1)-1 LOOP
+            FOR j IN i+1..array_length(col_num,1) LOOP
+                EXECUTE format(
+                  'INSERT INTO tmp_corr
+                   SELECT %L, %L, corr(%I,%I)
+                   FROM %I.%I',
+                   col_num[i], col_num[j], col_num[i], col_num[j], p_schema, p_table
+                );
+            END LOOP;
+        END LOOP;
+    END IF;
+
+    RAISE NOTICE '✅ Tablas temporales creadas. Revísalas en el navegador de DBeaver (prefijo tmp_)';
+END;
+
+
+LANGUAGE plpgsql;