Describe memory management (arena) + more complete parser + std.log (#8)

DESIGN.md

@@ -0,0 +1,33 @@
+# Design
+
+## Memory Management
+
+Authorizing a biscuit is a short-lived single pass operation. This makes arena allocation
+a great candidate for large parts of biscuit memory management:
+
+- It can greatly reduce the complexity of code that would otherwise have to carefully clone
+  / move memory to avoid leaks and double frees.
+- Potentially code can be faster because we don't do granular deallocation and we can avoid
+  some copying.
+- Again, because we're not necessarily copying, there is the potential for reduced memory usage
+  in places.
+
+The disadvantage would be that:
+
+- We potentially over allocate, i.e. we are storing more in memory than we technically need to.
+
+We create a toplevel arena and into it allocate all:
+
+- facts
+- predicates
+- rules
+- queries
+- policies
+- expressions
+- ops
+- terms
+
+i.e. all the domain level objects are arena allocated. This means we don't have to do
+complex reasoning about scope / lifetimes of these objects, they are all valid until
+the toplevel arena is deallocated. If we are careful to always copy when modifying
+one of these resources we can also share resources.

biscuit-builder/src/check.zig

@@ -5,25 +5,35 @@ const Term = @import("term.zig").Term;
 const Rule = @import("rule.zig").Rule;
 
 pub const Check = struct {
-    kind: datalog.Check.Kind,
+    kind: Kind,
     queries: std.ArrayList(Rule),
 
-    pub fn deinit(check: Check) void {
-        for (check.queries.items) |query| {
-            query.deinit();
-        }
+    pub const Kind = enum {
+        one,
+        all,
+    };
 
-        check.queries.deinit();
+    pub fn deinit(_: Check) void {
+        // for (check.queries.items) |query| {
+        //     query.deinit();
+        // }
+
+        // check.queries.deinit();
     }
 
-    pub fn convert(check: Check, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Check {
+    pub fn toDatalog(check: Check, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Check {
         var queries = std.ArrayList(datalog.Rule).init(allocator);
 
         for (check.queries.items) |query| {
-            try queries.append(try query.convert(allocator, symbols));
+            try queries.append(try query.toDatalog(allocator, symbols));
         }
 
-        return .{ .kind = check.kind, .queries = queries };
+        const kind: datalog.Check.Kind = switch (check.kind) {
+            .one => .one,
+            .all => .all,
+        };
+
+        return .{ .kind = kind, .queries = queries };
     }
 
     pub fn format(check: Check, comptime _: []const u8, _: std.fmt.FormatOptions, writer: anytype) !void {

biscuit-builder/src/expression.zig

@@ -58,7 +58,7 @@ pub const Expression = union(ExpressionType) {
     };
 
     /// convert to datalog fact
-    pub fn convert(expression: Expression, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Expression {
+    pub fn toDatalog(expression: Expression, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Expression {
         var ops = std.ArrayList(datalog.Op).init(allocator);
 
         try expression.toOpcodes(allocator, &ops, symbols);
@@ -68,7 +68,7 @@ pub const Expression = union(ExpressionType) {
 
     pub fn toOpcodes(expression: Expression, allocator: std.mem.Allocator, ops: *std.ArrayList(datalog.Op), symbols: *datalog.SymbolTable) !void {
         switch (expression) {
-            .value => |v| try ops.append(.{ .value = try v.convert(allocator, symbols) }),
+            .value => |v| try ops.append(.{ .value = try v.toDatalog(allocator, symbols) }),
             .unary => |u| {
                 try u.expression.toOpcodes(allocator, ops, symbols);
 
@@ -113,22 +113,22 @@ pub const Expression = union(ExpressionType) {
         }
     }
 
-    pub fn deinit(expression: *Expression) void {
-        switch (expression.*) {
-            .value => |v| v.deinit(),
-            .unary => |*u| {
-                u.expression.deinit();
-
-                u.allocator.destroy(u.expression);
-            },
-            .binary => |*b| {
-                b.left.deinit();
-                b.right.deinit();
-
-                b.allocator.destroy(b.left);
-                b.allocator.destroy(b.right);
-            },
-        }
+    pub fn deinit(_: *Expression) void {
+        // switch (expression.*) {
+        //     .value => |v| v.deinit(),
+        //     .unary => |*u| {
+        //         u.expression.deinit();
+
+        //         u.allocator.destroy(u.expression);
+        //     },
+        //     .binary => |*b| {
+        //         b.left.deinit();
+        //         b.right.deinit();
+
+        //         b.allocator.destroy(b.left);
+        //         b.allocator.destroy(b.right);
+        //     },
+        // }
     }
 
     pub fn value(term: Term) !Expression {
@@ -159,7 +159,7 @@ pub const Expression = union(ExpressionType) {
             .value => |v| try writer.print("{any}", .{v}),
             .unary => |u| {
                 switch (u.op) {
-                    .negate => try writer.print("-{any}", .{u.expression}),
+                    .negate => try writer.print("!{any}", .{u.expression}),
                     .parens => try writer.print("({any})", .{u.expression}),
                     .length => try writer.print("{any}.length()", .{u.expression}),
                 }

biscuit-builder/src/fact.zig

@@ -7,13 +7,13 @@ pub const Fact = struct {
     predicate: Predicate,
     variables: ?std.StringHashMap(?Term),
 
-    pub fn deinit(fact: Fact) void {
-        fact.predicate.deinit();
+    pub fn deinit(_: Fact) void {
+        // fact.predicate.deinit();
     }
 
     /// convert to datalog fact
-    pub fn convert(fact: Fact, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Fact {
-        return .{ .predicate = try fact.predicate.convert(allocator, symbols) };
+    pub fn toDatalog(fact: Fact, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Fact {
+        return .{ .predicate = try fact.predicate.toDatalog(allocator, symbols) };
     }
 
     pub fn format(fact: Fact, comptime _: []const u8, _: std.fmt.FormatOptions, writer: anytype) !void {

biscuit-builder/src/policy.zig

@@ -11,19 +11,19 @@ pub const Policy = struct {
         deny,
     };
 
-    pub fn deinit(policy: Policy) void {
-        for (policy.queries.items) |query| {
-            query.deinit();
-        }
+    pub fn deinit(_: Policy) void {
+        // for (policy.queries.items) |query| {
+        //     query.deinit();
+        // }
 
-        policy.queries.deinit();
+        // policy.queries.deinit();
     }
 
-    // pub fn convert(policy: Policy, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !Policy {
+    // pub fn toDatalog(policy: Policy, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !Policy {
     //     var queries = std.ArrayList(Rule).init(allocator);
 
     //     for (policy.queries.items) |query| {
-    //         try queries.append(try query.convert(allocator, symbols));
+    //         try queries.append(try query.toDatalog(allocator, symbols));
     //     }
 
     //     return .{ .kind = policy.kind, .queries = queries };

biscuit-builder/src/predicate.zig

@@ -6,22 +6,22 @@ pub const Predicate = struct {
     name: []const u8,
     terms: std.ArrayList(Term),
 
-    pub fn deinit(predicate: Predicate) void {
-        for (predicate.terms.items) |term| {
-            term.deinit();
-        }
+    pub fn deinit(_: Predicate) void {
+        // for (predicate.terms.items) |term| {
+        //     term.deinit();
+        // }
 
-        predicate.terms.deinit();
+        // predicate.terms.deinit();
     }
 
     /// convert to datalog predicate
-    pub fn convert(predicate: Predicate, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Predicate {
+    pub fn toDatalog(predicate: Predicate, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Predicate {
         const name = try symbols.insert(predicate.name);
 
         var terms = std.ArrayList(datalog.Term).init(allocator);
 
         for (predicate.terms.items) |term| {
-            try terms.append(try term.convert(allocator, symbols));
+            try terms.append(try term.toDatalog(allocator, symbols));
         }
 
         return .{ .name = name, .terms = terms };

biscuit-builder/src/root.zig

@@ -7,3 +7,4 @@ pub const Expression = @import("expression.zig").Expression;
 pub const Scope = @import("scope.zig").Scope;
 pub const Date = @import("date.zig").Date;
 pub const Policy = @import("policy.zig").Policy;
+pub const Set = @import("biscuit-datalog").Set;

biscuit-builder/src/rule.zig

@@ -12,40 +12,40 @@ pub const Rule = struct {
     variables: ?std.StringHashMap(?Term),
     scopes: std.ArrayList(Scope),
 
-    pub fn deinit(rule: Rule) void {
-        rule.head.deinit();
+    pub fn deinit(_: Rule) void {
+        // rule.head.deinit();
 
-        for (rule.body.items) |predicate| {
-            predicate.deinit();
-        }
+        // for (rule.body.items) |predicate| {
+        //     predicate.deinit();
+        // }
 
-        for (rule.expressions.items) |*expression| {
-            expression.deinit();
-        }
+        // for (rule.expressions.items) |*expression| {
+        //     expression.deinit();
+        // }
 
-        rule.body.deinit();
-        rule.expressions.deinit();
-        rule.scopes.deinit();
+        // rule.body.deinit();
+        // rule.expressions.deinit();
+        // rule.scopes.deinit();
     }
 
     /// convert to datalog predicate
-    pub fn convert(rule: Rule, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Rule {
-        const head = try rule.head.convert(allocator, symbols);
+    pub fn toDatalog(rule: Rule, allocator: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Rule {
+        const head = try rule.head.toDatalog(allocator, symbols);
 
         var body = std.ArrayList(datalog.Predicate).init(allocator);
         var expressions = std.ArrayList(datalog.Expression).init(allocator);
         var scopes = std.ArrayList(datalog.Scope).init(allocator);
 
         for (rule.body.items) |predicate| {
-            try body.append(try predicate.convert(allocator, symbols));
+            try body.append(try predicate.toDatalog(allocator, symbols));
         }
 
         for (rule.expressions.items) |expression| {
-            try expressions.append(try expression.convert(allocator, symbols));
+            try expressions.append(try expression.toDatalog(allocator, symbols));
         }
 
         for (rule.scopes.items) |scope| {
-            try scopes.append(try scope.convert(allocator, symbols));
+            try scopes.append(try scope.toDatalog(allocator, symbols));
         }
 
         return .{

biscuit-builder/src/scope.zig

@@ -19,7 +19,7 @@ pub const Scope = union(ScopeKind) {
     parameter: []const u8,
 
     /// convert to datalog fact
-    pub fn convert(scope: Scope, _: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Scope {
+    pub fn toDatalog(scope: Scope, _: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Scope {
         return switch (scope) {
             .authority => .{ .authority = {} },
             .previous => .{ .previous = {} },

biscuit-builder/src/term.zig

@@ -8,6 +8,8 @@ const TermTag = enum(u8) {
     integer,
     bool,
     date,
+    bytes,
+    set,
 };
 
 pub const Term = union(TermTag) {
@@ -16,17 +18,30 @@ pub const Term = union(TermTag) {
     integer: i64,
     bool: bool,
     date: u64,
+    bytes: []const u8,
+    set: datalog.Set(Term),
 
     pub fn deinit(_: Term) void {}
 
-    pub fn convert(term: Term, _: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Term {
-        return switch (term) {
-            .variable => |s| .{ .variable = @truncate(try symbols.insert(s)) }, // FIXME: assert symbol fits in u32
-            .string => |s| .{ .string = try symbols.insert(s) },
-            .integer => |n| .{ .integer = n },
-            .bool => |b| .{ .bool = b },
-            .date => |d| .{ .date = d },
-        };
+    pub fn toDatalog(term: Term, arena: std.mem.Allocator, symbols: *datalog.SymbolTable) !datalog.Term {
+        switch (term) {
+            .variable => |s| return .{ .variable = std.math.cast(u32, try symbols.insert(s)) orelse return error.FailedToCastInt },
+            .string => |s| return .{ .string = try symbols.insert(s) },
+            .integer => |n| return .{ .integer = n },
+            .bool => |b| return .{ .bool = b },
+            .date => |d| return .{ .date = d },
+            .bytes => |b| return .{ .bytes = b },
+            .set => |s| {
+                var datalog_set = datalog.Set(datalog.Term).init(arena);
+
+                var it = s.iterator();
+                while (it.next()) |t| {
+                    try datalog_set.add(try t.toDatalog(arena, symbols));
+                }
+
+                return .{ .set = datalog_set };
+            },
+        }
     }
 
     pub fn format(term: Term, comptime _: []const u8, _: std.fmt.FormatOptions, writer: anytype) !void {
@@ -36,6 +51,54 @@ pub const Term = union(TermTag) {
             .integer => |n| try writer.print("{}", .{n}),
             .bool => |b| if (b) try writer.print("true", .{}) else try writer.print("false", .{}),
             .date => |n| try writer.print("{}", .{n}),
+            .bytes => |b| try writer.print("{x}", .{b}),
+            .set => |s| {
+                try writer.print("[", .{});
+
+                const count = s.count();
+
+                var it = s.iterator();
+
+                var i: usize = 0;
+                while (it.next()) |t| {
+                    defer i += 1;
+                    try writer.print("{any}", .{t});
+
+                    if (i < count - 1) try writer.print(", ", .{});
+                }
+
+                try writer.print("]", .{});
+            },
+        }
+    }
+
+    pub fn eql(term: Term, other_term: Term) bool {
+        if (std.meta.activeTag(term) != std.meta.activeTag(other_term)) return false;
+
+        return switch (term) {
+            .variable => |v| std.mem.eql(u8, v, other_term.variable),
+            .integer => |v| v == other_term.integer,
+            .string => |v| std.mem.eql(u8, v, other_term.string),
+            .bool => |v| v == other_term.bool,
+            .date => |v| v == other_term.date,
+            .bytes => |v| std.mem.eql(u8, v, other_term.bytes),
+            .set => |v| v.eql(other_term.set),
+        };
+    }
+
+    pub fn hash(term: Term, hasher: anytype) void {
+        // Hash the tag type
+        std.hash.autoHash(hasher, std.meta.activeTag(term));
+
+        // Hash the value
+        switch (term) {
+            .variable => |v| for (v) |b| std.hash.autoHash(hasher, b),
+            .integer => |v| std.hash.autoHash(hasher, v),
+            .string => |v| for (v) |b| std.hash.autoHash(hasher, b),
+            .bool => |v| std.hash.autoHash(hasher, v),
+            .date => |v| std.hash.autoHash(hasher, v),
+            .bytes => |v| for (v) |b| std.hash.autoHash(hasher, b),
+            .set => |v| v.hash(hasher),
         }
     }
 };