From c1877f84f3596f67245abe6658b08c050bd1e686 Mon Sep 17 00:00:00 2001
From: Boris Kolpackov <boris@codesynthesis.com>
Date: Tue, 4 Sep 2012 11:22:07 +0200
Subject: NULL handling improvements

Add support for specifying NULL-ness for types with built-in mapping.
Handle Oracle [N]VARCHAR2 and SQLite FLOAT oddities using this mechanism
instead of overriding it at the schema generation level.

Also use the is_null argument that is passed to value_traits::init_image()
to indicate whether the value can be NULL.
---
 doc/manual.xhtml | 26 +++++++++++++++++---------
 1 file changed, 17 insertions(+), 9 deletions(-)

(limited to 'doc')

diff --git a/doc/manual.xhtml b/doc/manual.xhtml
index ab0ab65..a1e0360 100644
--- a/doc/manual.xhtml
+++ b/doc/manual.xhtml
@@ -13878,13 +13878,13 @@ class object
     <tr>
       <td><code>float</code></td>
       <td><code>REAL</code></td>
-      <td><code>NOT NULL</code></td>
+      <td><code>NULL</code></td>
     </tr>
 
     <tr>
       <td><code>double</code></td>
       <td><code>REAL</code></td>
-      <td><code>NOT NULL</code></td>
+      <td><code>NULL</code></td>
     </tr>
 
     <tr>
@@ -13894,6 +13894,14 @@ class object
     </tr>
   </table>
 
+  <p>SQLite represents <code>NaN</code> <code>FLOAT</code> values
+     as <code>NULL</code> values. As a result, columns of the
+     <code>float</code> and <code>double</code> types are by default
+     declared as <code>NULL</code>. However, you can override this by
+     explicitly declaring them as <code>NOT NULL</code> with the
+     <code>db&nbsp;not_null</code> pragma (<a href="#12.4.6">Section
+     12.4.6, "<code>null/not_null</code>"</a>).</p>
+
   <p>The SQLite ODB runtime library also provides support for mapping the
      <code>std::vector&lt;char></code>,
      <code>std::vector&lt;unsigned&nbsp;char></code>,
@@ -15344,13 +15352,13 @@ class object
 
   <p>In Oracle empty <code>VARCHAR2</code> and <code>NVARCHAR2</code>
      strings are represented as <code>NULL</code> values. As a result,
-     in the generated schema, columns of these types are declared as
-     <code>NULL</code> even if explicitly declared as
-     <code>NOT NULL</code> with the <code>db&nbsp;not_null</code> pragma
-     (<a href="#12.4.6">Section 12.4.6, "<code>null/not_null</code>"</a>),
-     except for primary key columns. This also means that for object ids
-     that are mapped to these Oracle types, an empty string is an invalid
-     value.</p>
+     columns of the <code>std::string</code> type are by default declared
+     as <code>NULL</code> except for primary key columns. However, you
+     can override this by explicitly declaring them as <code>NOT NULL</code>
+     with the <code>db&nbsp;not_null</code> pragma (<a href="#12.4.6">Section
+     12.4.6, "<code>null/not_null</code>"</a>). This also means that for
+     object ids that are mapped to these Oracle types, an empty string is
+     an invalid value.</p>
 
   <p>The Oracle ODB runtime library also provides support for mapping the
      <code>std::string</code> type to the Oracle <code>CHAR</code>,
-- 
cgit v1.1