joeyates/cpp-active-record

A C++ Implementation of the Active Record Pattern

cpp-active-record's Issues

Sort out test_helpers

Add tests for floating point

The fields() method should always return the same Fields no matter the context (ie. database fetch or class)

See #8 for context.

Use trailing '_' for member variables

Handle different query parameter placeholders

Placeholders are different between SQLite ('?') and PostgreSQL ('$1')

Add an example of parameter binding

Document how to use different types of database

SQLite - TableSet::update_table(...) always throwing (tries to drop primary key)

Reference:

table_set.cpp:68 
void TableSet::update_table( Table &required ) 

required.fields() and existing.fields() will always be different because of the automatic primary key assignment. Using the example table below:

td.fields() = fields
        ("code",        ActiveRecord::text)
        ("title",       ActiveRecord::text);

required.fields() will contain: [code, title]
existing.fields() will contain: [id, code, title]

Thus, the set difference required.fields() - existing.fields() will always return the primary key id and try to drop it.

Proposed solution:

void TableSet::update_table( Table &required ) {
  log( "TableSet::update_table" );
  log( required.table_name() );
  Table existing = table_data( required.connection(), required.table_name() );
  Fields missing = required.fields() - existing.fields();
  Fields remove  = existing.fields() - required.fields();
  for( Fields::iterator it = missing.begin(); it != missing.end(); ++it )
    existing.add_field( *it );
  for( Fields::iterator it = remove.begin(); it != remove.end(); ++it ) {
    if ((*it).name() == existing.primary_key()) /* fix */
      continue;
    throw ActiveRecordException( "Table::remove_field not yet implemented", __FILE__, __LINE__ );
    //existing.remove_field( *it );
  }
}

Better solution:

The fields() method should always return the same Fields no matter the context (ie. database fetch or class). The TableSet::table_data function should not load the primary key as a Field but instead only set the id_ variable of the Table. Of course I say this without fully understanding the impact, but hopefully you agree with keeping everything deterministic.

Use a linter

AR_HAS_MANY(...) macro generates compiler errors

When compiling, from the master branch, the unit tests fail when resolving and compiling the AR_HAS_MANY macro:

include/active_record/base.h:30:59: error: declaration of ‘std::vector<T1>    ActiveRecord::Base<T>::has_many()’ outside of class is not definition [-fpermissive]
 vector< other > ActiveRecord::Base< owner >::has_many(); \
                                                       ^
test/base_associations_test.cpp:29:1: note: in expansion of macro ‘AR_HAS_MANY’
 AR_HAS_MANY( Library, Book )

gcc version 4.8.2 (Ubuntu 4.8.2-19ubuntu1)

Standardise use of exceptions

Only raise exceptions which are subclasses of ActiveRecordException.

Put version in lib name

Unify code style

• whitespace

Add validations

Implement "timestamps" on TableData

If true, add 2 fields: created_at, updated_at

Uncaught exception

base.h
bool load();

stringstream is alive and used when throwing exception.
During throwing exception and memory deployment we have "Aborted".
This one works fine

template <class T>
bool Base<T>::load() {
  {
      std::stringstream ss;
      ss << "SELECT * ";
      ss << "FROM " << table_name_ << " ";
      ss << "WHERE ";
      ss << primary_key_ << " = ?";

      AttributeList parameters;
      parameters.push_back(id());

      Row row = connection_->select_one(ss.str(), parameters);
      if(row.has_data())
      {
          attributes_ = row.attributes();
          state_ = loaded;
          return true;
      }
  }
  throw ActiveRecordException("Record not found", __FILE__, __LINE__);
  return false;
}

Include configure script and Makefile in project

Using sqlite3 int64_t value type

Hello!
In your code of sqlite3_attribute.cpp we can see something like this:

// TYPE_LIST
  if(strcasecmp(type, "INTEGER") == 0) {
    return static_cast<int64_t>(sqlite3_column_int(pStmt, i));
  }

But if we ask https://www.sqlite.org we can get something like this:
SQLITE_API int sqlite3_column_int(sqlite3_stmt*, int iCol);
Of course we haven't int64_t becouse we have int
We can use something like this:
SQLITE_API sqlite3_int64 sqlite3_column_int64(sqlite3_stmt*, int iCol);
And finally:

// TYPE_LIST
  if(strcasecmp(type, "INTEGER") == 0) {
    return sqlite3_column_int64(pStmt, i);
  }

And also in sqlite3_connection.cpp

case Type::integer: {
        int64_t value = boost::get<int64_t>(parameter);
        pa.param_values[i] = nullptr;
        sqlite3_bind_int(ppStmt, i + 1, value);
        break;
      }

We must use sqlite3_bind_int64
Done)
This library is very useful for me. And I hope that we can make it even better.
Thanks)

Use transactions in generated code