Set version to 0.16

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = ProcruSQL
-version = 0.0.6
+version = 0.0.16
 author = Peter J. Holzer
 author_email = hjp@hjp.at
 description = Make a database fit its description
@@ -9,6 +9,7 @@ long_description_content_type = text/markdown
 url = https://git.hjp.at:3000/hjp/procrusql
 project_urls =
     Bug Tracker = https://git.hjp.at:3000/hjp/procrusql/issues
+    Repository = https://git.hjp.at:3000/hjp/procrusql
 classifiers =
     Programming Language :: Python :: 3
     License :: OSI Approved :: MIT License

src/procrusql/__init__.py

@@ -48,11 +48,12 @@ class Node:
         self.order = order
 
 class HaveData(Node):
-    def __init__(self, name, depends, table, key, extra):
+    def __init__(self, name, depends, table, key, extra, schema="public"):
         super().__init__(name, depends)
         self.table = table
         self.key = key
         self.extra = extra
+        self.schema = schema
 
     def check(self):
         log_check.info("Checking %s", self.name)
@@ -63,8 +64,9 @@ class HaveData(Node):
         ]
         key_check = sql.SQL(" and ").join(key_checks)
         q = sql.SQL(
-                "select * from {table} where {key_check}"
+                "select * from {schema}.{table} where {key_check}"
             ).format(
+                schema=sql.Identifier(self.schema),
                 table=sql.Identifier(self.table),
                 key_check=key_check
             )
@@ -73,40 +75,86 @@ class HaveData(Node):
         self.result = csr.fetchall()
         log_check.info("Got %d rows", len(self.result))
         if self.result:
+            extra_columns = list(self.extra.keys())
+            for c in extra_columns:
+                if self.result[0][c] != self.extra[c]:
+                    log_action.info("Updating %s: %s <- %s", key_values, c, self.extra[c])
+                    q = sql.SQL(
+                        "update {schema}.{table} set {column}={placeholder} where {key_check}"
+                    ).format(
+                        schema=sql.Identifier(self.schema),
+                        table=sql.Identifier(self.table),
+                        column=sql.Identifier(c),
+                        placeholder=sql.Placeholder(),
+                        key_check=key_check,
+                    )
+                    csr.execute(q, [self.extra[c]] + key_values)
+                    self.result[0][c] = self.extra[c]
             self.set_order()
             self.ok = True
             log_state.info("%s is now ok", self.name)
             return
-
-        extra_values = [v.resolve() if isinstance(v, Ref) else v for v in self.extra.values()]
-        columns = list(self.key.keys()) + list(self.extra.keys())
-        values = key_values + extra_values
-        q = sql.SQL(
-                "insert into {table}({columns}) values({placeholders}) returning *"
-            ).format(
-                table=sql.Identifier(self.table),
-                columns=sql.SQL(", ").join([sql.Identifier(x) for x in columns]),
-                placeholders=sql.SQL(", ").join([sql.Placeholder() for x in columns]),
-            )
-        log_action.info("Inserting data")
-        csr.execute(q, values)
-        self.result = csr.fetchall()
-        log_action.info("Got %d rows", len(self.result))
-        if self.result:
-            self.set_order()
-            self.ok = True
-            log_state.info("%s is now ok", self.name)
-            return
-        # We shouldn't get here. Either the insert succeeded, or it raised an
-        # exception. Success with 0 rows should not happen.
+        else:
+            extra_values = [v.resolve() if isinstance(v, Ref) else v for v in self.extra.values()]
+            columns = list(self.key.keys()) + list(self.extra.keys())
+            values = key_values + extra_values
+            q = sql.SQL(
+                    "insert into {schema}.{table}({columns}) values({placeholders}) returning *"
+                ).format(
+                    schema=sql.Identifier(self.schema),
+                    table=sql.Identifier(self.table),
+                    columns=sql.SQL(", ").join([sql.Identifier(x) for x in columns]),
+                    placeholders=sql.SQL(", ").join([sql.Placeholder() for x in columns]),
+                )
+            log_action.info("Inserting data")
+            csr.execute(q, values)
+            self.result = csr.fetchall()
+            log_action.info("Got %d rows", len(self.result))
+            if self.result:
+                self.set_order()
+                self.ok = True
+                log_state.info("%s is now ok", self.name)
+                return
+            # We shouldn't get here. Either the insert succeeded, or it raised an
+            # exception. Success with 0 rows should not happen.
         raise RuntimeError("Unreachable code reached")
 
+class HaveInit(Node):
+    def __init__(self, name, depends, table, query):
+        super().__init__(name, depends)
+        self.table = table
+        self.query = query
+
+    def check(self):
+        log_check.info("Checking %s", self.name)
+
+        for w in want:
+            if isinstance(w, HaveTable) and w.table == self.table:
+                if not w.ok:
+                    log_check.error("%s not yet ok", w.name)
+                    raise RuntimeError(f"Cannot insert into table {w.table} from {w.name} which is not yet ok. Please add a dependency")
+                if w.new:
+                    log_action.info("Executing %s", self.query)
+                    csr = db.cursor(cursor_factory=extras.DictCursor)
+                    csr.execute(self.query)
+                    self.result = csr.fetchall()
+                    log_action.info("Got %d rows", len(self.result))
+                    self.set_order()
+                else:
+                    log_check.info("Table %s already exists", w.table)
+                self.ok = True
+                log_state.info("%s is now ok", self.name)
+                break
+        else:
+            raise RuntimeError(f"Cannot find a rule which creates table {self.table}")
+
+
 class HaveTable(Node):
 
     def __init__(self, name, depends, table, schema="public"):
         super().__init__(name, depends)
         self.table = table
-        self.schema = "public"
+        self.schema = schema
 
     def check(self):
         log_check.info("Checking %s", self.name)
@@ -122,6 +170,7 @@ class HaveTable(Node):
             # Table exists, all ok
             self.set_order()
             self.ok = True
+            self.new = False
             log_state.info("%s is now ok", self.name)
             return
         if len(r) > 1:
@@ -137,6 +186,7 @@ class HaveTable(Node):
         csr.execute(q)
         self.set_order()
         self.ok = True
+        self.new = True
         log_state.info("%s is now ok", self.name)
 
     pass
@@ -152,7 +202,7 @@ class HaveColumn(Node):
         self.table = table
         self.column = column
         self.definition = definition
-        self.schema = "public"
+        self.schema = schema
 
     def check(self):
         log_check.info("Checking %s", self.name)
@@ -166,7 +216,18 @@ class HaveColumn(Node):
                 (self.schema, self.table, self.column, ))
         r = csr.fetchall()
         if len(r) == 1:
-            # Column exists, all ok
+            # Column exists, check attributes
+            if (r[0]["is_nullable"] == "YES") != self.definition["nullable"]:
+                log_action.info("Changing column %s of %s.%s to %s",
+                                self.column, self.schema, self.table,
+                                "null" if self.definition["nullable"] else "not null")
+                q = sql.SQL("alter table {schema}.{table} alter {column} {action} not null").format(
+                        schema=sql.Identifier(self.schema),
+                        table=sql.Identifier(self.table),
+                        column=sql.Identifier(self.column),
+                        action=sql.SQL("drop" if self.definition["nullable"] else "set")
+                    )
+                csr.execute(q)
             self.set_order()
             self.ok = True
             log_state.info("%s is now ok", self.name)
@@ -180,7 +241,7 @@ class HaveColumn(Node):
                 schema=sql.Identifier(self.schema),
                 table=sql.Identifier(self.table),
                 column=sql.Identifier(self.column),
-                definition=sql.SQL(self.definition),
+                definition=sql.SQL(self.definition["text"]),
             )
         csr.execute(q)
         self.set_order()
@@ -295,16 +356,22 @@ def main():
     ap = argparse.ArgumentParser()
     ap.add_argument("--dbname")
     ap.add_argument("--dbuser")
+    ap.add_argument("--schema", default="public")
     ap.add_argument("files", nargs="+")
     args = ap.parse_args()
 
     db = psycopg2.connect(dbname=args.dbname, user=args.dbuser)
+    csr = db.cursor()
+    csr.execute("show search_path")
+    search_path = csr.fetchone()[0]
+    search_path = args.schema + ", " + search_path
+    csr.execute(f"set search_path to {search_path}")
 
     rules = []
     for f in args.files:
         with open(f) as rf:
             text = rf.read()
-        ps = parser.ParseState(text)
+        ps = parser.ParseState(text, schema=args.schema)
 
         ps2 = parser.parse_ruleset(ps)
 

src/procrusql/parser.py

@@ -21,13 +21,14 @@ class Failure:
 
 class ParseState:
 
-    def __init__(self, text, position=0):
+    def __init__(self, text, position=0, schema="public"):
         self.text = text
         self.position = position
+        self.schema = schema # the default schema. probably doesn't belong into the parser state
         self.child_failure = None
 
     def clone(self):
-        ps = ParseState(self.text, self.position)
+        ps = ParseState(self.text, self.position, self.schema)
         return ps
 
     @property
@@ -42,7 +43,7 @@ class ParseState:
         linesbefore = self.text[:position].split("\n")
         linesafter = self.text[position:].split("\n")
         good = "\x1B[40;32m"
-        bad = "\x1B[40;31m"
+        bad = "\x1B[40;31;1m"
         reset = "\x1B[0m"
         s = reset + message + "\n"
         lines = []
@@ -78,7 +79,8 @@ def parse_ruleset(ps):
         ps3 = parse_table_rule(ps2) or  \
               parse_column_rule(ps2) or \
               parse_data_rule(ps2) or   \
-              parse_index_rule(ps2)
+              parse_index_rule(ps2) or  \
+              parse_view_rule(ps2)
         if ps3:
             ps2.ast.append(ps3.ast)
             ps2.position = ps3.position
@@ -98,8 +100,16 @@ def parse_table_rule(ps):
     if not ps3:
         ps.record_child_failure(ps2, "expected table name")
         return
+    if len(ps3.ast) == 2:
+        schema_name = ps3.ast[0]
+        table_name = ps3.ast[1]
+    elif len(ps3.ast) == 1:
+        schema_name = ps3.schema
+        table_name = ps3.ast[0]
+    else:
+        assert(False)
 
-    ps2.ast = procrusql.HaveTable(rulename(), [], ps3.ast[0])
+    ps2.ast = procrusql.HaveTable(rulename(), [], ps3.ast[0], schema=schema_name)
     ps2.position = ps3.position
     return ps2
 
@@ -119,7 +129,14 @@ def parse_column_rule(ps):
     if not ps3:
         ps.record_child_failure(ps2, "expected table name")
         return
-    table_name = ps3.ast[0]
+    if len(ps3.ast) == 2:
+        schema_name = ps3.ast[0]
+        table_name = ps3.ast[1]
+    elif len(ps3.ast) == 1:
+        schema_name = ps3.schema
+        table_name = ps3.ast[0]
+    else:
+        assert(False)
     ps2.position = ps3.position
 
     ps3 = parse_column_name(ps2)
@@ -136,7 +153,9 @@ def parse_column_rule(ps):
     column_definition = ps3.ast[0]
     ps2.position = ps3.position
 
-    ps2.ast = procrusql.HaveColumn(rulename(), [], table_name, column_name, column_definition)
+    ps2.ast = procrusql.HaveColumn(
+                            rulename(), [],
+                            table_name, column_name, column_definition, schema=schema_name)
 
     ps2.match_newlines()
 
@@ -153,33 +172,51 @@ def parse_data_rule(ps):
     if not ps3:
         ps.record_child_failure(ps2, "expected table name")
         return
-    table_name = ps3.ast[0]
-    ps2.position = ps3.position
-
-    ps3 = parse_dict(ps2)
-    if not ps3:
-        ps.record_child_failure(ps2, "expected key data definition")
-        return
-    key_data = ps3.ast
-    ps2.position = ps3.position
-
-    ps3 = parse_dict(ps2)
-    if not ps3:
-        ps.record_child_failure(ps2, "expected extra data definition")
-        return
-    extra_data = ps3.ast
-    ps2.position = ps3.position
-
-    ps3 = parse_label(ps2)
-    if ps3:
-        label = ps3.ast
-        ps2.position = ps3.position
+    if len(ps3.ast) == 2:
+        schema_name = ps3.ast[0]
+        table_name = ps3.ast[1]
+    elif len(ps3.ast) == 1:
+        schema_name = ps3.schema
+        table_name = ps3.ast[0]
     else:
+        assert(False)
+    ps2.position = ps3.position
+
+    if ps3 := parse_dict(ps2):
+        key_data = ps3.ast
+        ps2.position = ps3.position
+
+        ps3 = parse_dict(ps2)
+        if not ps3:
+            ps.record_child_failure(ps2, "expected extra data definition")
+            return
+        extra_data = ps3.ast
+        ps2.position = ps3.position
+
+        ps3 = parse_label(ps2)
+        if ps3:
+            label = ps3.ast
+            ps2.position = ps3.position
+        else:
+            label = rulename()
+
+        ps2.ast = procrusql.HaveData(
+                                label, [],
+                                table_name, key_data, extra_data, schema=schema_name)
+
+        ps2.match_newlines()
+    elif ps3 := parse_init_query(ps2):
+        # We have a bit of a problem here: The query extends to the end of the
+        # line so there is no room for a label. I really don't want to parse
+        # SQL here.
         label = rulename()
+        ps2.position = ps3.position
+        ps2.ast = procrusql.HaveInit(label, [], table_name, ps3.ast)
 
-    ps2.ast = procrusql.HaveData(label, [], table_name, key_data, extra_data)
-
-    ps2.match_newlines()
+        ps2.match_newlines()
+    else:
+        ps.record_child_failure(ps2, "expected key data definition or insert query")
+        return
 
     return ps2
 
@@ -206,7 +243,14 @@ def parse_index_rule(ps):
     if not ps3:
         ps.record_child_failure(ps2, "expected table name")
         return
-    table_name = ps3.ast[0]
+    if len(ps3.ast) == 2:
+        schema_name = ps3.ast[0]
+        table_name = ps3.ast[1]
+    elif len(ps3.ast) == 1:
+        schema_name = ps3.schema
+        table_name = ps3.ast[0]
+    else:
+        assert(False)
     ps2.position = ps3.position
 
     m = ps2.match(r"\s*(using\b|\([\w, ]+\))[^>\n]*")
@@ -221,12 +265,29 @@ def parse_index_rule(ps):
     else:
         label = rulename()
 
-    ps2.ast = procrusql.HaveIndex(label, [], table_name, index_name, index_type, index_definition)
+    ps2.ast = procrusql.HaveIndex(
+                            label, [],
+                            table_name, index_name,
+                            index_type, index_definition, schema=schema_name)
 
     ps2.match_newlines()
 
     return ps2
 
+def parse_view_rule(ps):
+    ps2 = ps.clone()
+    ps2.skip_whitespace_and_comments()
+    if not ps2.match(r"view\b"):
+        ps.record_child_failure(ps2, "expected “view”")
+        return
+    ps2.skip_whitespace_and_comments()
+    ps3 = parse_table_name(ps2)
+    if not ps3:
+        ps.record_child_failure(ps2, "expected view name")
+        return
+    ps2.skip_whitespace_and_comments()
+    ps3 = parse_multiline_string(ps2)
+
 
 
 def parse_table_name(ps):
@@ -238,6 +299,12 @@ def parse_table_name(ps):
     if ps2.rest[0].isalpha():
         m = ps2.match(r"\w+") # always succeeds since we already checked the first character
         ps2.ast.append(m.group(0))
+        if ps2.rest[0] == ".":
+            m = ps2.match(r"\w+")
+            if not m:
+                ps.record_child_failure(ps2, "expected table name after schema")
+                return
+            ps2.ast.append(m.group(0))
     else:
         ps.record_child_failure(ps2, "expected table name")
     return ps2
@@ -276,18 +343,30 @@ def parse_column_definition(ps):
     ps2 = ps.clone()
     ps2.ast = []
     ps2.skip_whitespace_and_comments()
-    sqltypes = (
-        "int", "integer", "serial",
-        "boolean",
-        "text", "character varying",
-        "date", "timestamp with time zone", "timestamptz"
+    sqltypes = sorted(
+        (
+            "integer", "int", "serial", "bigint",
+            "boolean",
+            "text", "character varying",
+            "date", "timestamp with time zone", "timestamptz",
+            "time",
+            "inet",
+            "double precision", "float8", "real", "float4",
+            "json", "jsonb",
+            "uuid",
+            r"integer\[\]", r"int\[\]", r"bigint\[\]",
+            "bytea",
+        ),
+        key=lambda x: -len(x) # longest match first
     )
-    pattern = "(" + "|".join(sqltypes) + ")" + r"([ \t]+default .*)?([ \t]+not null)?([ \t]+(primary key|unique|references \w+))?\b"
+    pattern = "(" + "|".join(sqltypes) + ")" + r"([ \t]+(default .*|not null\b|primary key\b|unique\b|references \w+\b))*"
     m = ps2.match(pattern)
     if not m:
         ps.record_child_failure(ps2, "expected column definition")
         return
-    ps2.ast.append(m.group(0))
+    text = m.group(0)
+    nullable = not("not null" in text or "primary key" in text)
+    ps2.ast.append({ "text": text, "nullable": nullable })
     return ps2
 
 def parse_dict(ps):
@@ -347,6 +426,16 @@ def parse_dict(ps):
     ps2.ast = d
     return ps2
 
+def parse_init_query(ps):
+    ps2 = ps.clone()
+    ps2.skip_whitespace_and_comments()
+    if m := ps2.match(r'(?i:with|insert)\b.*'):
+        ps2.ast = m.group(0)
+        return ps2
+    else:
+        ps.record_child_failure(ps2, "expected insert query")
+        return
+
 def parse_label(ps):
     ps2 = ps.clone()
     if m := ps2.match(r"\s*>>\s*(\w+)"):