第四部分 - 第一组测试（...以及 bug 们）

第三部分 - 一个只保存在内存中、只能新增记录的单表数据库

第五部分 - 持久化到磁盘

我们已经能够往我们的数据库中插入行和打印所有行了。花点时间来测试一下目前为止我们已经实现的功能。

因为我对 rspec 很熟悉，并且它的语法非常易读，因此我打算用它来编写我们的测试用例。

我将会定义一个简短的辅助函数来发送一系列指令到我们的数据库程序，并且对输出进行断言：

describe 'database' do
  def run_script(commands)
    raw_output = nil
    IO.popen("./db", "r+") do |pipe|
      commands.each do |command|
        pipe.puts command
      end

      pipe.close_write

      # 读取完整的输出
      raw_output = pipe.gets(nil)
    end
    raw_output.split("\n")
  end

  it 'inserts and retrieves a row' do
    result = run_script([
      "insert 1 user1 person1@example.com",
      "select",
      ".exit",
    ])
    expect(result).to match_array([
      "db > Executed.",
      "db > (1, user1, person1@example.com)",
      "Executed.",
      "db > ",
    ])
  end
end

这个简单的测试用例确保了我们可以从数据库取出我们放进去的数据，并且它也确实通过测试了：

bundle exec rspec
.

Finished in 0.00871 seconds (files took 0.09506 seconds to load)
1 example, 0 failures

那么现在测试插入大量数据到数据库变得可行了：

it 'prints error message when table is full' do
  script = (1..1401).map do |i|
    "insert #{i} user#{i} person#{i}@example.com"
  end
  script << ".exit"
  result = run_script(script)
  expect(result[-2]).to eq('db > Error: Table full.')
end

再次执行测试…

bundle exec rspec
..

Finished in 0.01553 seconds (files took 0.08156 seconds to load)
2 examples, 0 failures

测试通过了！我们的数据库可以保存 1400 行数据现在，因为我们设置了最大页数是 100，并且 14 行数据可以放在一页。

阅读现有的代码，我发现我们可能没办法正确保存文本类型的字段。通过下面这个简单的用例就可以测出来：

it 'allows inserting strings that are the maximum length' do
  long_username = "a"*32
  long_email = "a"*255
  script = [
    "insert 1 #{long_username} #{long_email}",
    "select",
    ".exit",
  ]
  result = run_script(script)
  expect(result).to match_array([
    "db > Executed.",
    "db > (1, #{long_username}, #{long_email})",
    "Executed.",
    "db > ",
  ])
end

测试没有通过！

Failures:

  1) database allows inserting strings that are the maximum length
     Failure/Error: raw_output.split("\n")

     ArgumentError:
       invalid byte sequence in UTF-8
     # ./spec/main_spec.rb:14:in `split'
     # ./spec/main_spec.rb:14:in `run_script'
     # ./spec/main_spec.rb:48:in `block (2 levels) in <top (required)>'

如果我们亲自试验一下，就会发现当我们打印这一行的时候会有一些奇奇怪怪的字符出现（省略了部分字符串）：

db > insert 1 aaaaa... aaaaa...
Executed.
db > select
(1, aaaaa...aaa\�, aaaaa...aaa\�)
Executed.
db >

发生了什么？如果你回过头看我们关于 Row 的定义，会发现我们只给 username 申请了刚好 32 字节空间、给 email 申请了刚好 255 字节空间。但是 C strings 是要以空字符作为结尾的，恰好我们没有为其申请空间。解决方案很简单，就是为它申请多一字节的空间：

 const uint32_t COLUMN_EMAIL_SIZE = 255;
 typedef struct {
   uint32_t id;
-  char username[COLUMN_USERNAME_SIZE];
-  char email[COLUMN_EMAIL_SIZE];
+  char username[COLUMN_USERNAME_SIZE + 1];
+  char email[COLUMN_EMAIL_SIZE + 1];
 } Row;

问题迎刃而解：

 bundle exec rspec
...

Finished in 0.0188 seconds (files took 0.08516 seconds to load)
3 examples, 0 failures

在插入数据时，我们不允许 username 和 email 的长度大于 32 和 255。那么我们的测试用例应该是类似这样的：

it 'prints error message if strings are too long' do
  long_username = "a"*33
  long_email = "a"*256
  script = [
    "insert 1 #{long_username} #{long_email}",
    "select",
    ".exit",
  ]
  result = run_script(script)
  expect(result).to match_array([
    "db > String is too long.",
    "db > Executed.",
    "db > ",
  ])
end

为了做到这点，需要修改下我们的解析器。提醒一下，目前输入我们用的是 scanf()：

if (strncmp(input_buffer->buffer, "insert", 6) == 0) {
  statement->type = STATEMENT_INSERT;
  int args_assigned = sscanf(
      input_buffer->buffer, "insert %d %s %s", &(statement->row_to_insert.id),
      statement->row_to_insert.username, statement->row_to_insert.email);
  if (args_assigned < 3) {
    return PREPARE_SYNTAX_ERROR;
  }
  return PREPARE_SUCCESS;
}

但是 scanf 有一些缺点。如果它读入的字符串长度超过了用于存放所读取的数据的内存空间，它将会造成溢出并且写入到一段不合法的内存区域。我们就需要在拷贝字符串到 Row 之前检查字符串的长度。为了实现这个目的，我们先将输入的内容按空格分隔。

我打算使用 strtok() （译者注：如果你不了解 strtok() 是做什么的，建议先点击链接阅读，这对你理解下面的代码会有很大帮助）来做字符串分割。我想你应该很容易看懂下面的代码：

+PrepareResult prepare_insert(InputBuffer* input_buffer, Statement* statement) {
+  statement->type = STATEMENT_INSERT;
+
+  char* keyword = strtok(input_buffer->buffer, " ");
+  char* id_string = strtok(NULL, " ");
+  char* username = strtok(NULL, " ");
+  char* email = strtok(NULL, " ");
+
+  if (id_string == NULL || username == NULL || email == NULL) {
+    return PREPARE_SYNTAX_ERROR;
+  }
+
+  int id = atoi(id_string);
+  if (strlen(username) > COLUMN_USERNAME_SIZE) {
+    return PREPARE_STRING_TOO_LONG;
+  }
+  if (strlen(email) > COLUMN_EMAIL_SIZE) {
+    return PREPARE_STRING_TOO_LONG;
+  }
+
+  statement->row_to_insert.id = id;
+  strcpy(statement->row_to_insert.username, username);
+  strcpy(statement->row_to_insert.email, email);
+
+  return PREPARE_SUCCESS;
+}
+
 PrepareResult prepare_statement(InputBuffer* input_buffer,
                                 Statement* statement) {
   if (strncmp(input_buffer->buffer, "insert", 6) == 0) {
+    return prepare_insert(input_buffer, statement);
-    statement->type = STATEMENT_INSERT;
-    int args_assigned = sscanf(
-        input_buffer->buffer, "insert %d %s %s", &(statement->row_to_insert.id),
-        statement->row_to_insert.username, statement->row_to_insert.email);
-    if (args_assigned < 3) {
-      return PREPARE_SYNTAX_ERROR;
-    }
-    return PREPARE_SUCCESS;
   }

对输入值连续调用 strtok 可以将其按照指定的分隔符（我们现在规定的是空格符）分割成好几个部分。strtok 的返回值是指向字符串的开始位置的指针。

然后我们就可以用 strlen() 来获取每个子字符串的长度。

同时加多一个错误码，用于表示字符串长度超过了预定的长度：

 enum PrepareResult_t {
   PREPARE_SUCCESS,
+  PREPARE_STRING_TOO_LONG,
   PREPARE_SYNTAX_ERROR,
   PREPARE_UNRECOGNIZED_STATEMENT
 };

 switch (prepare_statement(input_buffer, &statement)) {
   case (PREPARE_SUCCESS):
     break;
+  case (PREPARE_STRING_TOO_LONG):
+    printf("String is too long.\n");
+    continue;
   case (PREPARE_SYNTAX_ERROR):
     printf("Syntax error. Could not parse statement.\n");
     continue;

这样我们的测试就能通过了

bundle exec rspec
....

Finished in 0.02284 seconds (files took 0.116 seconds to load)
4 examples, 0 failures

到这里，我们可能还需要再处理一种错误的情况：

it 'prints an error message if id is negative' do
  script = [
    "insert -1 cstack foo@bar.com",
    "select",
    ".exit",
  ]
  result = run_script(script)
  expect(result).to match_array([
    "db > ID must be positive.",
    "db > Executed.",
    "db > ",
  ])
end

 enum PrepareResult_t {
   PREPARE_SUCCESS,
+  PREPARE_NEGATIVE_ID,
   PREPARE_STRING_TOO_LONG,
   PREPARE_SYNTAX_ERROR,
   PREPARE_UNRECOGNIZED_STATEMENT
@@ -148,9 +147,6 @@ PrepareResult prepare_insert(InputBuffer* input_buffer, Statement* statement) {
   }

   int id = atoi(id_string);
+  if (id < 0) {
+    return PREPARE_NEGATIVE_ID;
+  }
   if (strlen(username) > COLUMN_USERNAME_SIZE) {
     return PREPARE_STRING_TOO_LONG;
   }
@@ -230,9 +226,6 @@ int main(int argc, char* argv[]) {
     switch (prepare_statement(input_buffer, &statement)) {
       case (PREPARE_SUCCESS):
         break;
+      case (PREPARE_NEGATIVE_ID):
+        printf("ID must be positive.\n");
+        continue;
       case (PREPARE_STRING_TOO_LONG):
         printf("String is too long.\n");
         continue;

好了，目前为止我们的测试已经足够了。下一章将是一个非常重要的功能：持久化！我们将要把我们的数据保存到文件，并且还能将它读回到内存。

事情朝着好的方向发展了。

下面是本章的全部代码变更：

@@ -22,6 +22,8 @@

 enum PrepareResult_t {
   PREPARE_SUCCESS,
+  PREPARE_NEGATIVE_ID,
+  PREPARE_STRING_TOO_LONG,
   PREPARE_SYNTAX_ERROR,
   PREPARE_UNRECOGNIZED_STATEMENT
  };
@@ -34,8 +36,8 @@
 #define COLUMN_EMAIL_SIZE 255
 typedef struct {
   uint32_t id;
-  char username[COLUMN_USERNAME_SIZE];
-  char email[COLUMN_EMAIL_SIZE];
+  char username[COLUMN_USERNAME_SIZE + 1];
+  char email[COLUMN_EMAIL_SIZE + 1];
 } Row;

@@ -150,18 +152,40 @@ MetaCommandResult do_meta_command(InputBuffer* input_buffer, Table *table) {
   }
 }

-PrepareResult prepare_statement(InputBuffer* input_buffer,
-                                Statement* statement) {
-  if (strncmp(input_buffer->buffer, "insert", 6) == 0) {
+PrepareResult prepare_insert(InputBuffer* input_buffer, Statement* statement) {
   statement->type = STATEMENT_INSERT;
-  int args_assigned = sscanf(
-     input_buffer->buffer, "insert %d %s %s", &(statement->row_to_insert.id),
-     statement->row_to_insert.username, statement->row_to_insert.email
-     );
-  if (args_assigned < 3) {
+
+  char* keyword = strtok(input_buffer->buffer, " ");
+  char* id_string = strtok(NULL, " ");
+  char* username = strtok(NULL, " ");
+  char* email = strtok(NULL, " ");
+
+  if (id_string == NULL || username == NULL || email == NULL) {
      return PREPARE_SYNTAX_ERROR;
   }
+
+  int id = atoi(id_string);
+  if (id < 0) {
+     return PREPARE_NEGATIVE_ID;
+  }
+  if (strlen(username) > COLUMN_USERNAME_SIZE) {
+     return PREPARE_STRING_TOO_LONG;
+  }
+  if (strlen(email) > COLUMN_EMAIL_SIZE) {
+     return PREPARE_STRING_TOO_LONG;
+  }
+
+  statement->row_to_insert.id = id;
+  strcpy(statement->row_to_insert.username, username);
+  strcpy(statement->row_to_insert.email, email);
+
   return PREPARE_SUCCESS;
+
+}
+PrepareResult prepare_statement(InputBuffer* input_buffer,
+                                Statement* statement) {
+  if (strncmp(input_buffer->buffer, "insert", 6) == 0) {
+      return prepare_insert(input_buffer, statement);
   }
   if (strcmp(input_buffer->buffer, "select") == 0) {
     statement->type = STATEMENT_SELECT;
@@ -223,6 +247,12 @@ int main(int argc, char* argv[]) {
     switch (prepare_statement(input_buffer, &statement)) {
       case (PREPARE_SUCCESS):
         break;
+      case (PREPARE_NEGATIVE_ID):
+	printf("ID must be positive.\n");
+	continue;
+      case (PREPARE_STRING_TOO_LONG):
+	printf("String is too long.\n");
+	continue;
       case (PREPARE_SYNTAX_ERROR):
 	printf("Syntax error. Could not parse statement.\n");
 	continue;

测试用例：

+describe 'database' do
+  def run_script(commands)
+    raw_output = nil
+    IO.popen("./db", "r+") do |pipe|
+      commands.each do |command|
+        pipe.puts command
+      end
+
+      pipe.close_write
+
+      # Read entire output
+      raw_output = pipe.gets(nil)
+    end
+    raw_output.split("\n")
+  end
+
+  it 'inserts and retrieves a row' do
+    result = run_script([
+      "insert 1 user1 person1@example.com",
+      "select",
+      ".exit",
+    ])
+    expect(result).to match_array([
+      "db > Executed.",
+      "db > (1, user1, person1@example.com)",
+      "Executed.",
+      "db > ",
+    ])
+  end
+
+  it 'prints error message when table is full' do
+    script = (1..1401).map do |i|
+      "insert #{i} user#{i} person#{i}@example.com"
+    end
+    script << ".exit"
+    result = run_script(script)
+    expect(result[-2]).to eq('db > Error: Table full.')
+  end
+
+  it 'allows inserting strings that are the maximum length' do
+    long_username = "a"*32
+    long_email = "a"*255
+    script = [
+      "insert 1 #{long_username} #{long_email}",
+      "select",
+      ".exit",
+    ]
+    result = run_script(script)
+    expect(result).to match_array([
+      "db > Executed.",
+      "db > (1, #{long_username}, #{long_email})",
+      "Executed.",
+      "db > ",
+    ])
+  end
+
+  it 'prints error message if strings are too long' do
+    long_username = "a"*33
+    long_email = "a"*256
+    script = [
+      "insert 1 #{long_username} #{long_email}",
+      "select",
+      ".exit",
+    ]
+    result = run_script(script)
+    expect(result).to match_array([
+      "db > String is too long.",
+      "db > Executed.",
+      "db > ",
+    ])
+  end
+
+  it 'prints an error message if id is negative' do
+    script = [
+      "insert -1 cstack foo@bar.com",
+      "select",
+      ".exit",
+    ]
+    result = run_script(script)
+    expect(result).to match_array([
+      "db > ID must be positive.",
+      "db > Executed.",
+      "db > ",
+    ])
+  end
+end

第三部分 - 一个只保存在内存中、只能新增记录的单表数据库

第五部分 - 持久化到磁盘