Make Crawl open sqlite dbs using readonly connections where possible, and don't create SQLite transactions unless updating the db.

Crawl can now correctly handle the case where one or more Crawl processes are
running, the database source text files are modified, and a new Crawl is
started. The new Crawl will update the db with the new data, which the old
Crawl processes will also see immediately.

1 files changed, 10 insertions, 1 deletions

diff --git a/crawl-ref/source/sqldbm.h b/crawl-ref/source/sqldbm.h
index 15123b4945..35cff068d0 100644
--- a/crawl-ref/source/sqldbm.h
+++ b/crawl-ref/source/sqldbm.h
@@ -49,7 +49,8 @@ private:
 class SQL_DBM
 {
 public:
-    SQL_DBM(const std::string &db = "", bool open = false);
+    SQL_DBM(const std::string &db = "", bool readonly = true,
+            bool open = false);
     ~SQL_DBM();
 
     bool is_open() const;
@@ -62,6 +63,7 @@ public:
 
     std::string query(const std::string &key);
     int insert(const std::string &key, const std::string &value);
+    int remove(const std::string &key);
 
 public:
     std::string error;
@@ -73,15 +75,22 @@ private:
     int init_query();
     int init_iterator();
     int init_insert();
+    int init_remove();
     int init_schema();
     int ec(int err);
 
+    int try_insert(const std::string &key, const std::string &value);
+    int do_insert(const std::string &key, const std::string &value);
+    int do_query(const std::string &key, std::string *result);
+
 private:
     sqlite3      *db;
     sqlite3_stmt *s_insert;
+    sqlite3_stmt *s_remove;
     sqlite3_stmt *s_query;
     sqlite3_stmt *s_iterator;
     std::string  dbfile;
+    bool readonly;
 };
 
 SQL_DBM  *dbm_open(const char *filename, int open_mode, int permissions);