sequel_pg 1.5.1-x86-mingw32 → 1.6.1-x86-mingw32

data/CHANGELOG

@@ -1,3 +1,11 @@
+=== 1.6.1 (2012-10-25)
+
+* Make PostgreSQL array parser handle string encodings correctly on ruby 1.9 (jeremyevans)
+
+=== 1.6.0 (2012-09-04)
+
+* Replace PQsetRowProcessor streaming with PQsetSingleRowMode streaming introduced in PostgreSQL 9.2beta3 (jeremyevans)
+
 === 1.5.1 (2012-08-02)
 
 * Sprinkle some RB_GC_GUARD to work around segfaults in the PostgreSQL array parser (jeremyevans)

data/README.rdoc

@@ -65,7 +65,7 @@ enable the model optimization via:
 
 == Streaming
 
-If you are using PostgreSQL 9.2 or higher on the client, then sequel_pg
+If you are using PostgreSQL 9.2beta3 or higher on the client, then sequel_pg
 should enable streaming support.  This allows you to stream returned
 rows one at a time, instead of collecting the entire result set in
 memory (which is how PostgreSQL works by default).  You can check
@@ -76,8 +76,7 @@ if streaming is supported by:
 If streaming is supported, you can load the streaming support into the
 database:
 
-  require 'sequel_pg/streaming'
-  DB.extend Sequel::Postgres::Streaming
+  DB.extension(:pg_streaming)
 
 Then you can call the Dataset#stream method to have the dataset use
 the streaming support:
@@ -87,7 +86,11 @@ the streaming support:
 If you want to enable streaming for all of a database's datasets, you
 can do the following:
 
-  DB.extend_datasets Sequel::Postgres::Streaming::AllQueries
+  DB.stream_all_queries = true
+
+Note that pg 0.14.1+ is required for streaming to work.  This is not
+required by the gem, as it is only a requirement for streaming, not
+for general use.
 
 == Installing the gem
 

data/ext/sequel_pg/extconf.rb

@@ -14,7 +14,7 @@ if enable_config("static-build")
 end
 
 if (have_library('pq') || have_library('libpq') || have_library('ms/libpq')) && have_header('libpq-fe.h')
-  have_func 'PQsetRowProcessor'
+  have_func 'PQsetSingleRowMode'
   create_makefile("sequel_pg")
 else
   puts 'Could not find PostgreSQL build environment (libraries & headers): Makefile not created'

data/ext/sequel_pg/sequel_pg.c

@@ -103,6 +103,12 @@ static ID spg_id_columns;
 static ID spg_id_encoding;
 static ID spg_id_values;
 
+#if HAVE_PQSETSINGLEROWMODE
+static ID spg_id_get_result;
+static ID spg_id_clear;
+static ID spg_id_check;
+#endif
+
 #if SPG_ENCODING
 static int enc_get_index(VALUE val)
 {
@@ -114,7 +120,11 @@ static int enc_get_index(VALUE val)
 }
 #endif
 
-static VALUE read_array(int *index, char *c_pg_array_string, int array_string_length, char *word, VALUE converter)
+static VALUE read_array(int *index, char *c_pg_array_string, int array_string_length, char *word, VALUE converter
+#ifdef SPG_ENCODING
+, int enc_index
+#endif
+)
 {
   int word_index = 0;
 
@@ -155,15 +165,20 @@ static VALUE read_array(int *index, char *c_pg_array_string, int array_string_le
           {
             rb_ary_push(array, Qnil);
           }
-          else if (RTEST(converter))
+          else 
           {
-            VALUE rword = rb_str_new(word, word_index);
+            VALUE rword = rb_tainted_str_new(word, word_index);
             RB_GC_GUARD(rword);
-            rb_ary_push(array, rb_funcall(converter, spg_id_call, 1, rword));
-          }
-          else
-          {
-            rb_ary_push(array, rb_str_new(word, word_index));
+
+#ifdef SPG_ENCODING
+            rb_enc_associate_index(rword, enc_index);
+#endif
+
+            if (RTEST(converter)) {
+              rword = rb_funcall(converter, spg_id_call, 1, rword);
+            }
+
+            rb_ary_push(array, rword);
           }
         }
         if(c == '}')
@@ -181,7 +196,11 @@ static VALUE read_array(int *index, char *c_pg_array_string, int array_string_le
       else if(c == '{')
       {
         (*index)++;
-        rb_ary_push(array, read_array(index, c_pg_array_string, array_string_length, word, converter));
+        rb_ary_push(array, read_array(index, c_pg_array_string, array_string_length, word, converter
+#ifdef SPG_ENCODING
+, enc_index
+#endif
+        ));
         escapeNext = 1;
       }
       else
@@ -224,7 +243,11 @@ static VALUE parse_pg_array(VALUE self, VALUE pg_array_string, VALUE converter)
   char *word = RSTRING_PTR(buf);
   int index = 1;
 
-  return read_array(&index, c_pg_array_string, array_string_length, word, converter);
+  return read_array(&index, c_pg_array_string, array_string_length, word, converter
+#ifdef SPG_ENCODING
+, enc_get_index(pg_array_string)
+#endif
+  );
 }
 
 static VALUE spg_time(const char *s) {
@@ -403,7 +426,7 @@ static VALUE spg_timestamp(const char *s, VALUE self) {
 }
 
 static VALUE spg_fetch_rows_set_cols(VALUE self, VALUE ignore) {
-  return self;
+  return Qnil;
 }
 
 static VALUE spg__col_value(VALUE self, PGresult *res, long i, long j, VALUE* colconvert
@@ -813,194 +836,115 @@ static VALUE spg_yield_hash_rows(VALUE self, VALUE rres, VALUE ignore) {
 
 static VALUE spg_supports_streaming_p(VALUE self) {
   return
-#if HAVE_PQSETROWPROCESSOR
+#if HAVE_PQSETSINGLEROWMODE
   Qtrue;
 #else
   Qfalse;
 #endif
 }
 
-#if HAVE_PQSETROWPROCESSOR
-static VALUE spg__rp_value(VALUE self, PGresult* res, const PGdataValue* dvs, int j, VALUE* colconvert
-#ifdef SPG_ENCODING
-, int enc_index
-#endif
-) {
-  const char *v;
-  PGdataValue dv = dvs[j];
-  VALUE rv;
-  size_t l;
-  int len = dv.len;
-
-  if(len < 0) {
-    rv = Qnil;
-  } else {
-    v = dv.value;
-
-    switch(PQftype(res, j)) {
-      case 16: /* boolean */
-        rv = *v == 't' ? Qtrue : Qfalse;
-        break;
-      case 17: /* bytea */
-        v = PQunescapeBytea((unsigned char*)v, &l);
-        rv = rb_funcall(spg_Blob, spg_id_new, 1, rb_str_new(v, l));
-        PQfreemem((char *)v);
-        break;
-      case 20: /* integer */
-      case 21:
-      case 22:
-      case 23:
-      case 26:
-        rv = rb_str2inum(rb_str_new(v, len), 10);
-        break;
-      case 700: /* float */
-      case 701:
-        if (strncmp("NaN", v, 3) == 0) {
-          rv = spg_nan;
-        } else if (strncmp("Infinity", v, 8) == 0) {
-          rv = spg_pos_inf;
-        } else if (strncmp("-Infinity", v, 9) == 0) {
-          rv = spg_neg_inf;
-        } else {
-          rv = rb_float_new(rb_str_to_dbl(rb_str_new(v, len), Qfalse));
-        }
-        break;
-      case 790: /* numeric */
-      case 1700:
-        rv = rb_funcall(spg_BigDecimal, spg_id_new, 1, rb_str_new(v, len));
-        break;
-      case 1082: /* date */
-        rv = rb_str_new(v, len);
-        rv = spg_date(StringValuePtr(rv));
-        break;
-      case 1083: /* time */
-      case 1266:
-        rv = rb_str_new(v, len);
-        rv = spg_time(StringValuePtr(rv));
-        break;
-      case 1114: /* timestamp */
-      case 1184:
-        rv = rb_str_new(v, len);
-        rv = spg_timestamp(StringValuePtr(rv), self);
-        break;
-      case 18: /* char */
-      case 25: /* text */
-      case 1043: /* varchar*/
-        rv = rb_tainted_str_new(v, len);
-#ifdef SPG_ENCODING
-        rb_enc_associate_index(rv, enc_index);
-#endif
-        break;
-      default:
-        rv = rb_tainted_str_new(v, len);
-#ifdef SPG_ENCODING
-        rb_enc_associate_index(rv, enc_index);
-#endif
-        if (colconvert[j] != Qnil) {
-          rv = rb_funcall(colconvert[j], spg_id_call, 1, rv); 
-        }
-    }
+#if HAVE_PQSETSINGLEROWMODE
+static VALUE spg_set_single_row_mode(VALUE self) {
+  PGconn *conn;
+  Data_Get_Struct(self, PGconn, conn);
+  if (PQsetSingleRowMode(conn) != 1) {
+      rb_raise(spg_PGError, "cannot set single row mode");
   }
-  return rv;
+  return Qnil;
 }
 
-static int spg_row_processor(PGresult *res, const PGdataValue *columns, const char **errmsgp, void *param) {
+static VALUE spg__yield_each_row(VALUE self) {
+  PGconn *conn;
+  PGresult *res;
+  VALUE rres;
+  VALUE rconn;
+  VALUE colsyms[SPG_MAX_FIELDS];
+  VALUE colconvert[SPG_MAX_FIELDS];
   long nfields;
-  struct spg_row_proc_info *info;
-  info = (struct spg_row_proc_info *)param;
-  VALUE *colsyms = info->colsyms;
-  VALUE *colconvert = info->colconvert;
-  VALUE self = info->dataset;
+  long j;
+  VALUE h;
+  VALUE opts;
+  VALUE pg_type;
+  VALUE pg_value = Qnil;
+  char type = SPG_YIELD_NORMAL;
 
-  switch (PQresultStatus(res))
-  {
-    case PGRES_TUPLES_OK:
-    case PGRES_COPY_OUT:
-    case PGRES_COPY_IN:
-#ifdef HAVE_CONST_PGRES_COPY_BOTH
-    case PGRES_COPY_BOTH:
+  rconn = rb_ary_entry(self, 1);
+  self = rb_ary_entry(self, 0);
+  Data_Get_Struct(rconn, PGconn, conn);
+
+  rres = rb_funcall(rconn, spg_id_get_result, 0);
+  rb_funcall(rres, spg_id_check, 0);
+  Data_Get_Struct(rres, PGresult, res);
+
+#ifdef SPG_ENCODING
+  int enc_index;
+  enc_index = enc_get_index(rres);
 #endif
-    case PGRES_EMPTY_QUERY:
-    case PGRES_COMMAND_OK:
-      break;
-    case PGRES_BAD_RESPONSE:
-    case PGRES_FATAL_ERROR:
-    case PGRES_NONFATAL_ERROR:
-      rb_raise(spg_PGError, "error while streaming results");
-    default:
-      rb_raise(spg_PGError, "unexpected result status while streaming results");
+
+  /* Only handle regular and model types.  All other types require compiling all
+   * of the results at once, which is not a use case for streaming.  The streaming
+   * code does not call this function for the other types. */
+  opts = rb_funcall(self, spg_id_opts, 0);
+  if (rb_type(opts) == T_HASH) {
+    pg_type = rb_hash_aref(opts, spg_sym__sequel_pg_type);
+    pg_value = rb_hash_aref(opts, spg_sym__sequel_pg_value);
+    if (SYMBOL_P(pg_type) && pg_type == spg_sym_model && rb_type(pg_value) == T_CLASS) {
+      type = SPG_YIELD_MODEL;
+    }
   }
 
   nfields = PQnfields(res);
-  if(columns == NULL) {
-    spg_set_column_info(self, res, colsyms, colconvert);
-    rb_ivar_set(self, spg_id_columns, rb_ary_new4(nfields, colsyms));
-  } else {
-    long j;
-    VALUE h, m;
-    h = rb_hash_new();
+  if (nfields > SPG_MAX_FIELDS) {
+    rb_funcall(rres, spg_id_clear, 0);
+    rb_raise(rb_eRangeError, "more than %d columns in query", SPG_MAX_FIELDS);
+  }
 
+  spg_set_column_info(self, res, colsyms, colconvert);
+
+  rb_ivar_set(self, spg_id_columns, rb_ary_new4(nfields, colsyms));
+
+  while (PQntuples(res) != 0) {
+    h = rb_hash_new();
     for(j=0; j<nfields; j++) {
-      rb_hash_aset(h, colsyms[j], spg__rp_value(self, res, columns, j, colconvert
-#ifdef SPG_ENCODING
-                   , info->enc_index
-#endif
-      ));
+      rb_hash_aset(h, colsyms[j], spg__col_value(self, res, 0, j, colconvert ENC_INDEX));
     }
 
-    /* optimize_model_load used, return model instance */
-    if ((m = info->model)) {
-      m = rb_obj_alloc(m);
-      rb_ivar_set(m, spg_id_values, h);
-      h = m;
+    rb_funcall(rres, spg_id_clear, 0);
+
+    if(type == SPG_YIELD_MODEL) {
+      /* Abuse local variable */
+      pg_type = rb_obj_alloc(pg_value);
+      rb_ivar_set(pg_type, spg_id_values, h);
+      rb_yield(pg_type);
+    } else {
+      rb_yield(h);
     }
 
-    rb_funcall(info->block, spg_id_call, 1, h);
+    rres = rb_funcall(rconn, spg_id_get_result, 0);
+    rb_funcall(rres, spg_id_check, 0);
+    Data_Get_Struct(rres, PGresult, res);
   }
-  return 1;
-}
+  rb_funcall(rres, spg_id_clear, 0);
 
-static VALUE spg_unset_row_processor(VALUE rconn) {
-  PGconn *conn;
-  Data_Get_Struct(rconn, PGconn, conn);
-  if ((PQskipResult(conn)) != NULL) {
-    /* Results remaining when row processor finished,
-     * either because an exception was raised or the iterator
-     * exited early, so skip all remaining rows. */
-    while(PQgetResult(conn) != NULL) {
-      /* Use a separate while loop as PQgetResult is faster than
-       * PQskipResult. */
-    }
-  }
-  PQsetRowProcessor(conn, NULL, NULL);
-  return Qnil;
+  return self;
 }
 
-static VALUE spg_with_row_processor(VALUE self, VALUE rconn, VALUE dataset, VALUE block) {
-  struct spg_row_proc_info info;
+static VALUE spg__flush_results(VALUE rconn) {
   PGconn *conn;
+  PGresult *res;
   Data_Get_Struct(rconn, PGconn, conn);
-  bzero(&info, sizeof(info));
 
-  info.dataset = dataset;
-  info.block = block;
-  info.model = 0;
-#if SPG_ENCODING
-  info.enc_index = enc_get_index(rconn);
-#endif
-
-  /* Abuse local variable, detect if optimize_model_load used */
-  block = rb_funcall(dataset, spg_id_opts, 0);
-  if (rb_type(block) == T_HASH && rb_hash_aref(block, spg_sym__sequel_pg_type) == spg_sym_model) {
-    block = rb_hash_aref(block, spg_sym__sequel_pg_value);
-    if (rb_type(block) == T_CLASS) {
-      info.model = block;
-    }
+  while ((res = PQgetResult(conn)) != NULL) {
+    PQclear(res);
   }
 
-  PQsetRowProcessor(conn, spg_row_processor, (void*)&info);
-  rb_ensure(rb_yield, Qnil, spg_unset_row_processor, rconn);
-  return Qnil;
+  return rconn;
+}
+
+static VALUE spg_yield_each_row(VALUE self, VALUE rconn) {
+  VALUE v;
+  v = rb_ary_new3(2, self, rconn);
+  return rb_ensure(spg__yield_each_row, v, spg__flush_results, rconn);
 }
 #endif
 
@@ -1081,9 +1025,14 @@ void Init_sequel_pg(void) {
 
   rb_define_singleton_method(spg_Postgres, "supports_streaming?", spg_supports_streaming_p, 0);
 
-#if HAVE_PQSETROWPROCESSOR
-  c = rb_funcall(spg_Postgres, cg, 1, rb_str_new2("Database"));
-  rb_define_private_method(c, "with_row_processor", spg_with_row_processor, 3);
+#if HAVE_PQSETSINGLEROWMODE
+  spg_id_get_result = rb_intern("get_result");
+  spg_id_clear = rb_intern("clear");
+  spg_id_check = rb_intern("check");
+
+  rb_define_private_method(c, "yield_each_row", spg_yield_each_row, 1);
+  c = rb_funcall(spg_Postgres, cg, 1, rb_str_new2("Adapter"));
+  rb_define_private_method(c, "set_single_row_mode", spg_set_single_row_mode, 0);
 #endif
 
   rb_define_singleton_method(spg_Postgres, "parse_pg_array", parse_pg_array, 2);

data/lib/sequel/extensions/pg_streaming.rb

@@ -0,0 +1,130 @@
+unless Sequel::Postgres.respond_to?(:supports_streaming?)
+  raise LoadError, "either sequel_pg not loaded, or an old version of sequel_pg loaded"
+end
+unless Sequel::Postgres.supports_streaming?
+  raise LoadError, "streaming is not supported by the version of libpq in use"
+end
+
+# Database methods necessary to support streaming.  You should load this extension
+# into your database object:
+#
+#   DB.extension(:pg_streaming)
+#
+# Then you can call #stream on your datasets to use the streaming support:
+#
+#   DB[:table].stream.each{|row| ...}
+#
+# Or change a set so that all dataset calls use streaming:
+#
+#   DB.stream_all_queries = true
+module Sequel::Postgres::Streaming
+  attr_accessor :stream_all_queries
+
+  # Also extend the database's datasets to support streaming.
+  # This extension requires modifying connections, so disconnect
+  # so that new connections will get the methods.
+  def self.extended(db)
+    db.extend_datasets(DatasetMethods)
+    db.stream_all_queries = false
+    db.disconnect
+  end
+
+  # Make sure all new connections have the appropriate methods added.
+  def connect(server)
+    conn = super
+    conn.extend(AdapterMethods)
+    conn
+  end
+
+  private
+
+  # If streaming is requested, and a prepared statement is not
+  # used, tell the connection to use single row mode for the query.
+  def _execute(conn, sql, opts={}, &block)
+    if opts[:stream] && !sql.is_a?(Symbol) 
+      conn.single_row_mode = true
+    end
+    super
+  end
+
+  # If streaming is requested, send the prepared statement instead
+  # of executing it and blocking.
+  def _execute_prepared_statement(conn, ps_name, args, opts)
+    if opts[:stream]
+      conn.send_prepared_statement(ps_name, args)
+    else
+      super
+    end
+  end
+
+  module AdapterMethods
+    # Whether the next query on this connection should use
+    # single_row_mode.
+    attr_accessor :single_row_mode
+
+    # Send the prepared statement on this connection using
+    # single row mode.
+    def send_prepared_statement(ps_name, args)
+      send_query_prepared(ps_name, args)
+      set_single_row_mode
+      block
+      self
+    end
+
+    private
+
+    # If using single row mode, send the query instead of executing it.
+    def execute_query(sql, args)
+      if @single_row_mode
+        @single_row_mode = false
+        @db.log_yield(sql, args){args ? send_query(sql, args) : send_query(sql)}
+        set_single_row_mode
+        block
+        self
+      else
+        super
+      end
+    end
+  end
+
+  # Dataset methods used to implement streaming.
+  module DatasetMethods
+    # If streaming has been requested and the current dataset
+    # can be streamed, request the database use streaming when
+    # executing this query, and use yield_each_row to process
+    # the separate PGresult for each row in the connection.
+    def fetch_rows(sql)
+      if stream_results?
+        execute(sql, :stream=>true) do |conn|
+          yield_each_row(conn){|h| yield h}
+        end
+      else
+        super
+      end
+    end
+
+    # Return a clone of the dataset that will use streaming to load
+    # rows.
+    def stream
+      clone(:stream=>true)
+    end
+
+    private
+
+    # Only stream results if streaming has been specifically requested
+    # and the query is streamable.
+    def stream_results?
+      (@opts[:stream] || db.stream_all_queries) && streamable?
+    end
+
+    # Queries using cursors are not streamable, and queries that use
+    # the map/select_map/to_hash/to_hash_groups optimizations are not
+    # streamable, but other queries are streamable.
+    def streamable?
+      spgt = (o = @opts)[:_sequel_pg_type]
+      (spgt.nil? || spgt == :model) && !o[:cursor]
+    end
+  end
+end
+
+Sequel::Database.register_extension(:pg_streaming, Sequel::Postgres::Streaming)

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification 
 name: sequel_pg
 version: !ruby/object:Gem::Version 
-  hash: 1
+  hash: 13
   prerelease: 
   segments: 
   - 1
-  - 5
+  - 6
   - 1
-  version: 1.5.1
+  version: 1.6.1
 platform: x86-mingw32
 authors: 
 - Jeremy Evans
@@ -15,7 +15,7 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2012-04-07 00:00:00 Z
+date: 2012-09-28 00:00:00 Z
 dependencies: 
 - !ruby/object:Gem::Dependency 
   name: pg
@@ -41,12 +41,12 @@ dependencies:
     requirements: 
     - - ">="
       - !ruby/object:Gem::Version 
-        hash: 151
+        hash: 155
         segments: 
         - 3
-        - 36
+        - 39
         - 0
-        version: 3.36.0
+        version: 3.39.0
   type: :runtime
   version_requirements: *id002
 description: |
@@ -72,7 +72,7 @@ files:
 - ext/sequel_pg/extconf.rb
 - ext/sequel_pg/sequel_pg.c
 - lib/sequel_pg/sequel_pg.rb
-- lib/sequel_pg/streaming.rb
+- lib/sequel/extensions/pg_streaming.rb
 - lib/1.8/sequel_pg.so
 - lib/1.9/sequel_pg.so
 homepage: http://github.com/jeremyevans/sequel_pg

data/lib/sequel_pg/streaming.rb

@@ -1,82 +0,0 @@
-unless Sequel::Postgres.respond_to?(:supports_streaming?)
-  raise LoadError, "either sequel_pg not loaded, or an old version of sequel_pg loaded"
-end
-unless Sequel::Postgres.supports_streaming?
-  raise LoadError, "streaming is not supported by the version of libpq in use"
-end
-
-# Database methods necessary to support streaming.  You should extend your
-# Database object with this:
-#
-#   DB.extend Sequel::Postgres::Streaming
-#
-# Then you can call #stream on your datasets to use the streaming support:
-#
-#  DB[:table].stream.each{|row| ...}
-module Sequel::Postgres::Streaming
-  # Also extend the database's datasets to support streaming
-  def self.extended(db)
-    db.extend_datasets(DatasetMethods)
-  end
-
-  private
-
-  # If streaming is requested, set a row processor while executing
-  # the query.
-  def _execute(conn, sql, opts={})
-    if stream = opts[:stream]
-      with_row_processor(conn, *stream){super}
-    else
-      super
-    end
-  end
-
-  # Dataset methods used to implement streaming.
-  module DatasetMethods
-    # If streaming has been requested and the current dataset
-    # can be streamed, request the database use streaming when
-    # executing this query.
-    def fetch_rows(sql, &block)
-      if stream_results?
-        execute(sql, :stream=>[self, block])
-      else
-        super
-      end
-    end
-
-    # Return a clone of the dataset that will use streaming to load
-    # rows.
-    def stream
-      clone(:stream=>true)
-    end
-
-    private
-
-    # Only stream results if streaming has been specifically requested
-    # and the query is streamable.
-    def stream_results?
-      @opts[:stream] && streamable?
-    end
-
-    # Queries using cursors are not streamable, and queries that use
-    # the map/select_map/to_hash/to_hash_groups optimizations are not
-    # streamable, but other queries are streamable.
-    def streamable?
-      spgt = (o = @opts)[:_sequel_pg_type]
-      (spgt.nil? || spgt == :model) && !o[:cursor]
-    end
-  end
-
-  # Extend a database's datasets with this module to enable streaming
-  # on all streamable queries:
-  #
-  #   DB.extend_datasets(Sequel::Postgres::Streaming::AllQueries)
-  module AllQueries
-    private
-
-    # Always stream results if the query is streamable.
-    def stream_results?
-      streamable?
-    end
-  end
-end