Merge remote-tracking branch 'origin/main' into 3562-tablet-client-lo…

…adtable-should-show-error-msg

.github/workflows/other-os-build.yml

@@ -78,41 +78,60 @@ jobs:
           shell: bash
           run: |
             cd /root/OpenMLDB
+            # centos6_build.sh will try build zetasql even cache hit, just ignore the failure
             IN_WORKFLOW=true bash steps/centos6_build.sh
             # bazel bin
             export PATH=$PATH:`pwd`
             source /opt/rh/devtoolset-8/enable
             if [[ "${USE_DEPS_CACHE}" != "true" ]]; then
-              echo "build thirdparty"
-              make thirdparty CMAKE_INSTALL_PREFIX=${OPENMLDB_PREFIX} BUILD_BUNDLE=ON NPROC=8
+              echo "build thirdparty, make opt is better than nproc?"
+              make thirdparty CMAKE_INSTALL_PREFIX=${OPENMLDB_PREFIX} BUILD_BUNDLE=ON THIRD_PARTY_CMAKE_FLAGS=-DMAKEOPTS=-j8
+              # 5.8G	./.deps, avail 8G
               rm -rf .deps/build # GitHub runner disk space is limited
             fi
             echo "build"
+            # 1.4G	./.deps, avail 13G
+
+            # will failed if openmldb_sdk is on
             cmake -S . -B `pwd`/build -DCMAKE_PREFIX_PATH=`pwd`/.deps/usr -DCMAKE_BUILD_TYPE=RelWithDebInfo \
             -DSQL_PYSDK_ENABLE=${SQL_PYSDK_ENABLE} -DSQL_JAVASDK_ENABLE=OFF \
             -DTESTING_ENABLE=OFF -DCMAKE_INSTALL_PREFIX=${OPENMLDB_PREFIX}  \
             -DHYBRIDSE_TESTING_ENABLE=OFF -DEXAMPLES_ENABLE=OFF -DEXAMPLES_TESTING_ENABLE=OFF
-            cmake --build build --target install -- -j2
-            # clean up to save disk space(~11G), don't know which is relative, build again in next step
-            rm -rf build
+            # target openmldb 6.7G    ./build(no py/java), avail 5.2G
+            # openmldb+cp_python_sdk_so 7.7G    ./build(has py), python just ~180M
+            # target 'install' cost more, preinstall/fast won't build all, so use install/fast if needed
+            # or https://cmake.org/cmake/help/latest/variable/CMAKE_SKIP_INSTALL_ALL_DEPENDENCY.html
+            cmake --build build --target openmldb cp_python_sdk_so -- -j2
+            du -h --max-depth=1
+            df -h
+            # if target above cost too much disk, make java build failed, try to rm build cache
+            # don't rm cache now cuz build java from emtpy will cost 20min
+            # rm build/hybridse build/src -rf
             if [[ "${SQL_JAVASDK_ENABLE}" == "ON" ]]; then
               echo "build java sdk"
               cmake -S . -B `pwd`/build -DCMAKE_PREFIX_PATH=`pwd`/.deps/usr -DCMAKE_BUILD_TYPE=RelWithDebInfo \
               -DSQL_PYSDK_ENABLE=OFF -DSQL_JAVASDK_ENABLE=ON \
               -DTESTING_ENABLE=OFF -DCMAKE_INSTALL_PREFIX=${OPENMLDB_PREFIX}  \
               -DHYBRIDSE_TESTING_ENABLE=OFF -DEXAMPLES_ENABLE=OFF -DEXAMPLES_TESTING_ENABLE=OFF
-              cmake --build build --target sql_javasdk_package -- -j2
+              # if build the whole java, 7.6G	./build, 5.7G  ./java, avail 331M
+              # so split it and build native only
+              # 7.6G	./build,  1.8G	./java, avail 5.2G
+              cmake --build build --target cp_native_so -- -j2
+              du -h --max-depth=1
+              df -h
+              rm build/hybridse build/src -rf
+              cd java
+              ./mvnw -pl openmldb-native clean package -DskipTests=true -Dscalatest.skip=true -Dwagon.skip=true -Dmaven.test.skip=true --batch-mode 
             fi
-
-      - name: package    
-        run: |
-          tar czf ${{ env.OPENMLDB_PREFIX }}.tar.gz ${{ env.OPENMLDB_PREFIX }}/
+            rm build/hybridse build/src -rf
+            du -h --max-depth=1
+            df -h
 
       - name: upload binary
         uses: actions/upload-artifact@v2
         with:
-          path: openmldb-*.tar.gz
-          name: binary-package
+          path: build/bin/openmldb
+          name: binary
 
       - name: upload java native
         if: ${{ env.SQL_JAVASDK_ENABLE == 'ON' }}
@@ -127,8 +146,7 @@ jobs:
         with: 
           name: python-whl
           path: |
-            python/openmldb_sdk/dist/openmldb*.whl  
-    
+            python/openmldb_sdk/dist/openmldb*.whl
       # TODO(hw): upload cxx sdk
 
   # macos no need to build thirdparty, but binary/os needs to be built on each os

.github/workflows/sdk.yml

@@ -68,8 +68,6 @@ jobs:
         with:
           path: ~/.m2/repository
           key: ${{ runner.os }}-maven-${{ hashFiles('java/**/pom.xml') }}
-          restore-keys: |
-            ${{ runner.os }}-maven-
 
       - name: prepare release
         if: github.event_name == 'push'
@@ -124,6 +122,7 @@ jobs:
       - name: maven coverage
         working-directory: java
         run: |
+          rm -rfv ~/.m2/repository/com/4paradigm/
           ./mvnw --batch-mode prepare-package
           ./mvnw --batch-mode scoverage:report
 
@@ -160,8 +159,6 @@ jobs:
         with:
           path: ~/.m2/repository
           key: ${{ runner.os }}-maven-${{ hashFiles('java/**/pom.xml') }}
-          restore-keys: |
-            ${{ runner.os }}-maven-
 
       - name: Cache thirdparty
         uses: actions/cache@v3
@@ -236,6 +233,10 @@ jobs:
           MAVEN_USERNAME: ${{ secrets.OSSRH_USERNAME }}
           MAVEN_TOKEN: ${{ secrets.OSSRH_TOKEN }}
           GPG_PASSPHRASE: ${{ secrets.GPG_PASSPHRASE }}
+      - name: cleanup
+        run: |
+          rm -rfv ~/.m2/repository/com/4paradigm/
+
 
   python-sdk:
     runs-on: ubuntu-latest

.gitignore

@@ -108,3 +108,10 @@ allure-results/
 /python/openmldb_autofe/*.egg-info/
 # go sdk
 !go.mod
+
+# tag files
+**/tags
+**/GPATH
+**/GRTAGS
+**/GTAGS
+**/cscope.out

CMakeLists.txt

@@ -36,7 +36,7 @@ endif()
 message (STATUS "CMAKE_BUILD_TYPE: ${CMAKE_BUILD_TYPE}")
 set(OPENMLDB_VERSION_MAJOR 0)
 set(OPENMLDB_VERSION_MINOR 8)
-set(OPENMLDB_VERSION_BUG 2)
+set(OPENMLDB_VERSION_BUG 3)
 
 function(get_commitid CODE_DIR COMMIT_ID)
     find_package(Git REQUIRED)

Makefile

@@ -139,34 +139,44 @@ THIRD_PARTY_BUILD_DIR ?= $(MAKEFILE_DIR)/.deps
 THIRD_PARTY_SRC_DIR ?= $(MAKEFILE_DIR)/thirdsrc
 THIRD_PARTY_DIR ?= $(THIRD_PARTY_BUILD_DIR)/usr
 
-# trick: for those compile inside hybridsql docker image, thirdparty is pre-installed in /deps/usr.
-#  we check this by asserting if the environment variable '$THIRD_PARTY_DIR' is defined to '/deps/usr',
-#  if true, thirdparty download is skipped
-# zetasql check separately since it update more frequently:
-#  it will updated if the variable '$ZETASQL_VERSION' (defined in docker) not equal to that defined in current code
-override GREP_PATTERN = "set(ZETASQL_VERSION"
+override ZETASQL_PATTERN = "set(ZETASQL_VERSION"
+override THIRD_PATTERN = "set(HYBRIDSQL_ASSERTS_VERSION"
+new_zetasql_version := $(shell grep $(ZETASQL_PATTERN) third-party/cmake/FetchZetasql.cmake | sed 's/[^0-9.]*\([0-9.]*\).*/\1/')
+new_third_version := $(shell grep $(THIRD_PATTERN) third-party/CMakeLists.txt | sed 's/[^0-9.]*\([0-9.]*\).*/\1/')
+
 thirdparty-fast:
 	@if [ $(THIRD_PARTY_DIR) != "/deps/usr" ] ; then \
 	    echo "[deps]: install thirdparty and zetasql"; \
 	    $(MAKE) thirdparty; \
-	elif [ -n "$(ZETASQL_VERSION)" ]; then \
-	    new_zetasql_version=$(shell grep $(GREP_PATTERN) third-party/cmake/FetchZetasql.cmake | sed 's/[^0-9.]*\([0-9.]*\).*/\1/'); \
-	    if [ "$$new_zetasql_version" != "$(ZETASQL_VERSION)" ] ; then \
-		echo "[deps]: thirdparty up-to-date. reinstall zetasql from $(ZETASQL_VERSION) to $$new_zetasql_version"; \
-		$(MAKE) thirdparty-configure; \
-		$(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR) -j $(NPROC) --target zetasql; \
-	    else \
-		echo "[deps]: all up-to-date. zetasql already installed with version: $(ZETASQL_VERSION)"; \
-	    fi; \
 	else \
-	    echo "[deps]: install zetasql only"; \
 	    $(MAKE) thirdparty-configure; \
-	    $(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR) --target zetasql; \
+	    if [ -n "$(ZETASQL_VERSION)" ] ; then \
+		if [ "$(new_zetasql_version)" != "$(ZETASQL_VERSION)" ] ; then \
+		    echo "[deps]: installing zetasql from $(ZETASQL_VERSION) to $(new_zetasql_version)"; \
+		    $(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR) --target zetasql; \
+		else \
+		    echo "[deps]: zetasql up-to-date with version: $(ZETASQL_VERSION)"; \
+		fi; \
+	    else \
+		echo "[deps]: installing latest zetasql"; \
+		$(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR) --target zetasql; \
+	    fi;  \
+	    if [ -n "$(THIRDPARTY_VERSION)" ]; then \
+		if [ "$(new_third_version)" != "$(THIRDPARTY_VERSION)" ] ; then \
+		    echo "[deps]: installing thirdparty from $(THIRDPARTY_VERSION) to $(new_third_version)"; \
+		    $(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR) --target hybridsql-asserts; \
+		else \
+		    echo "[deps]: thirdparty up-to-date: $(THIRDPARTY_VERSION)"; \
+		fi ; \
+	    else \
+		echo "[deps]: installing latest thirdparty"; \
+		$(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR) --target hybridsql-asserts; \
+	    fi ; \
 	fi
 
 # third party compiled code install to 'OpenMLDB/.deps/usr', source code install to 'OpenMLDB/thirdsrc'
 thirdparty: thirdparty-configure
-	$(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR) -j $(NPROC)
+	$(CMAKE_PRG) --build $(THIRD_PARTY_BUILD_DIR)
 
 thirdparty-configure:
 	$(CMAKE_PRG) -S third-party -B $(THIRD_PARTY_BUILD_DIR) -DSRC_INSTALL_DIR=$(THIRD_PARTY_SRC_DIR) -DDEPS_INSTALL_DIR=$(THIRD_PARTY_DIR) $(THIRD_PARTY_CMAKE_FLAGS)

cases/function/window/error_window.yaml

@@ -17,15 +17,17 @@ debugs: []
 version: 0.5.0
 cases:
   - id: 0
-    desc: no order by
+    desc: RANGE-type WINDOW with offset PRECEDING/FOLLOWING requires ORDER BY
     inputs:
       - columns: [ "id int","c1 string","c3 int","c4 bigint","c5 float","c6 double","c7 timestamp","c8 date" ]
         indexs: [ "index1:c8:c4" ]
         rows:
           - [1,"aa",20,30,1.1,2.1,1590738990000,"2020-05-01"]
     sql: |
-      SELECT id, c1, c4, count(c4) OVER w1 as w1_c4_count FROM {0} WINDOW w1 AS (PARTITION BY {0}.c8 ROWS BETWEEN 2 PRECEDING AND CURRENT ROW);
+      SELECT id, c1, c4, count(c4) OVER w1 as w1_c4_count FROM {0}
+      WINDOW w1 AS (PARTITION BY {0}.c8 ROWS_RANGE BETWEEN 2 PRECEDING AND CURRENT ROW);
     expect:
+      msg: RANGE/ROWS_RANGE-type FRAME with offset PRECEDING/FOLLOWING requires exactly one ORDER BY column
       success: false
   - id: 1
     desc: no partition by
@@ -301,3 +303,29 @@ cases:
       SELECT id, c1, c3, sum(c4) OVER w1 as w1_c4_sum FROM {0} WINDOW w1 AS (PARTITION BY {0}.c33 ORDER BY {0}.c7 ROWS_RANGE BETWEEN 2s PRECEDING AND CURRENT ROW);
     expect:
       success: false
+  - id: 17
+    desc: ROWS WINDOW + EXCLUDE CURRENT_TIME requires order by
+    inputs:
+      - columns: [ "id int","c1 string","c3 int","c4 bigint","c5 float","c6 double","c7 timestamp","c8 date" ]
+        indexs: [ "index1:c8:c4" ]
+        rows:
+          - [1,"aa",20,30,1.1,2.1,1590738990000,"2020-05-01"]
+    sql: |
+      SELECT id, c1, c4, count(c4) OVER w1 as w1_c4_count FROM {0}
+      WINDOW w1 AS (PARTITION BY {0}.c8 ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW EXCLUDE CURRENT_TIME);
+    expect:
+      msg: WINDOW with EXCLUDE CURRENT_TIME requires exactly one ORDER BY column
+      success: false
+  - id: 18
+    desc: RANGE WINDOW + EXCLUDE CURRENT_TIME requires order by
+    inputs:
+      - columns: [ "id int","c1 string","c3 int","c4 bigint","c5 float","c6 double","c7 timestamp","c8 date" ]
+        indexs: [ "index1:c8:c4" ]
+        rows:
+          - [1,"aa",20,30,1.1,2.1,1590738990000,"2020-05-01"]
+    sql: |
+      SELECT id, c1, c4, count(c4) OVER w1 as w1_c4_count FROM {0}
+      WINDOW w1 AS (PARTITION BY {0}.c8 ROWS_RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW EXCLUDE CURRENT_TIME);
+    expect:
+      msg: WINDOW with EXCLUDE CURRENT_TIME requires exactly one ORDER BY column
+      success: false

cases/plan/cmd.yaml

@@ -704,3 +704,20 @@ cases:
           +-actions:
             +-0: DropPathAction (12)
             +-1: AddPathAction (13)
+
+  - id: show-create-table
+    desc: SHOW CREATE TABLE
+    sql: SHOW CREATE TABLE t1;
+    expect:
+      node_tree_str: |
+        +-node[CMD]
+          +-cmd_type: show create table
+          +-args: [t1]
+  - id: show-create-table-db
+    desc: SHOW CREATE TABLE
+    sql: SHOW CREATE TABLE db1.t1;
+    expect:
+      node_tree_str: |
+        +-node[CMD]
+          +-cmd_type: show create table
+          +-args: [db1, t1]

cases/query/const_query.yaml

@@ -126,3 +126,55 @@ cases:
       columns: ['c1 bool', 'c2 int16', 'c3 int', 'c4 double', 'c5 string', 'c6 date', 'c7 timestamp' ]
       rows:
         - [ true, 3, 13, 10.0, 'a string', '2020-05-22', 1590115420000 ]
+
+  # =================================================================================
+  # Null safe for structure types: String, Date, Timestamp and Array
+  # creating struct from:
+  # 1. NULL liternal (const null)
+  # 2. another supported date type but fails to cast, e.g. timestamp(-1) returns NULL of timestamp
+  #
+  # casting to array type un-implemented
+  # =================================================================================
+  - id: 10
+    desc: null safe for date
+    mode: procedure-unsupport
+    sql: |
+      select
+        datediff(Date(timestamp(-1)), Date("2021-05-01")) as out1,
+        datediff(Date(timestamp(-2177481600)), Date("2021-05-01")) as out2,
+        datediff(cast(NULL as date), Date("2021-05-01")) as out3,
+        date(NULL) as out4,
+        date("abc") as out5,
+        date(timestamp("abc")) as out6
+    expect:
+      columns: ["out1 int", "out2 int", "out3 int", "out4 date", "out5 date", "out6 date"]
+      data: |
+        NULL, NULL, NULL, NULL, NULL, NULL
+  - id: 11
+    desc: null safe for timestamp
+    mode: procedure-unsupport
+    sql: |
+      select
+        month(cast(NULL as timestamp)) as out1,
+        month(timestamp(NULL)) as out2,
+        month(timestamp(-1)) as out3,
+        month(timestamp("abc")) as out4,
+        month(timestamp(date("abc"))) as out5
+    expect:
+      columns: ["out1 int", "out2 int", "out3 int", "out4 int", "out5 int"]
+      data: |
+        NULL, NULL, NULL, NULL, NULL
+  - id: 12
+    desc: null safe for string
+    mode: procedure-unsupport
+    sql: |
+      select
+        char_length(cast(NULL as string)) as out1,
+        char_length(string(int(NULL))) as out2,
+        char_length(string(bool(null))) as out3,
+        char_length(string(timestamp(null))) as out4,
+        char_length(string(date(null))) as out5
+    expect:
+      columns: ["out1 int", "out2 int", "out3 int", "out4 int", "out5 int"]
+      data: |
+        NULL, NULL, NULL, NULL, NULL