schemas/util.sql: cluster(): support omitting index to recluster a table

schemas/util.sql: cluster_once(): use util.cluster()

schemas/util.sql: cluster_once(): documented that this runs cluster only once if called repeatedly

schemas/util.sql: added cluster(), which reclusters each time

schemas/util.sql: cluster_once(): use simpler `LANGUAGE sql` now that we're using util.eval() instead of EXECUTE

schemas/util.sql: cluster_once(): use util.eval() instead of EXECUTE to print the CLUSTER statement (if any) it's using

bugfix: schemas/util.sql: cluster_once(): need to use just the name of the index, without the schema prefix

schemas/util.sql: cluster_once(): use simpler IS DISTINCT FROM instead of array !=

bugfix: schemas/util.sql: view_is_subset(): auto-updatable views that are not plain subsets: include ones with WHERE clauses

bugfix: schemas/util.sql: view_is_subset(): also need to handle auto-updatable views that are not plain subsets (there are some of these)

bugfix: schemas/util.sql: comment(element oid): need `objsubid = 0` filter so this doesn't use comments for any other objsubids

bugfix: schemas/util.sql: rematerialize_view(): also need to handle the target table's dependent views

bugfix: schemas/util.sql: mk_drop_from_create(): need to use `IF EXISTS` so that the resulting statement also works if the target object does not yet exist

schemas/util.sql: mk_drop_from_create(): also support `SELECT util.drop_*()` statements, in addition to DROP

schemas/util.sql: replace_words(): use map_words() to enable inlining the concatenated hstore, so that it doesn't need to be concatenated on each function call

schemas/util.sql: added map_words()

schemas/util.sql: added date_from_spanish()

schemas/util.sql: added replace_words()

schemas/util.sql: added words(text)

schemas/util.sql: spanish_date_words(): switched to using hstore because this is better for the algorithm that will translate these

schemas/util.sql: added spanish_date_words()

schemas/util.sql: added `replacement` type

bugfix: mk_set_relation_metadata(): need to include col comments as well so that these are re-created properly by recreate()

schemas/util.sql: added mk_set_comments(regclass)

schemas/util.sql: show_set_comment(regclass): renamed to mk_set_comment() to match other mk_set_comment() functions

schemas/util.sql: added mk_set_col_comments(regclass)

schemas/util.sql: added cols(regclass)

schemas/util.sql: added mk_set_comment(col_ref)

schemas/util.sql: mk_set_comment(table_ regclass, comment text): use new util.mk_set_comment(text, text)

schemas/util.sql: added mk_set_comment(col col_ref, comment text)

schemas/util.sql: added sql(col_ref)

schemas/util.sql: added mk_set_comment(on_ text, comment text)

schemas/util.sql: added comment(col_ref)

schemas/util.sql: added col_num(col_ref)

schemas/util.sql: in_south_america(): optimized by using BETWEEN instead of util.contained_within__no_dateline(). this reduces the filter time for 2014-6-4.Iara_Lacher.reserve_prioritization by about 10%.

bugfix: in_south_america(): must use util.contained_within__no_dateline() instead of util.contained_within_approx() to ensure that the more accurate geometry logic is used

schemas/util.sql: materialize_view(): use util.copy() instead of util.materialize_query() so that all view metadata is transferred

schemas/util.sql: added subspecies()

schemas/util.sql: added south_america(), in_south_america()

fix: schemas/util.sql: OPERATOR ~(geocoord, postgis.geometry): renamed to because this is not an approximate comparison for geometry

fix: schemas/util.sql: contained_within_approx(geocoord, postgis.geometry): renamed to contained_within__no_dateline(__) because this is not an approximate comparison for geometry

schemas/util.sql: lat_long_in_new_world(): renamed to just in_new_world() because the lat/long is implied by the param type

schemas/util.sql: lat_long_in_new_world(): take a geocoord param instead of separate lat/long params

schemas/util.sql: added contained_within_approx(geocoord, geometry) and corresponding OPERATOR ~@(geocoord, geometry)

schemas/util.sql: added OPERATOR ~@(geocoord, geography)

schemas/util.sql: lat_long_in_new_world(): use new contained_within_approx(geocoord, geography)

schemas/util.sql: added contained_within_approx(geocoord, postgis.geography), which enables specifying just `(lat, long)` without the ::util.geocoord type specifier

schemas/util.sql: OPERATOR (postgis.geography, postgis.geography): renamed to ~ because it's approximate

schemas/util.sql: contained_within(): renamed to contained_within_approx() because the latitude lines of geography type bounding boxes bulge outward, creating false positives above and below the bounding box

schemas/util.sql: added contained_within__no_dateline(geometry, geometry) and corresponding operator @

schemas/util.sql: contained_within(): renamed to contained_within_approx() because the latitude lines of geography type bounding boxes bulge outward, creating false positives above and below the bounding box

schemas/util.sql: geometry(geocoord): documented that it is not possible to create a cast for this, as a bug in pg_dump prevents the cast from being exported, even when no export filters are applied

schemas/util.sql: point(geocoord): renamed to geometry(geocoord) since this is now a cast

schemas/util.sql: point(): return geometry instead of geography to support using points with geometry arithmetic

schemas/util.sql: point(): take a single util.geocoord param instead of separate lat/long

schemas/util.sql: added geocoord type

schemas/util.sql: bounding_box(): use bounding_box__no_dateline() to construct the postgis.geometry object

schemas/util.sql: added bounding_box__no_dateline(), which is more accurate than util.bounding_box() (latitude lines will be straight), but geocoordinate wraparound is not supported

bugfix: schemas/util.sql: view_def_to_orig(): need to handle cases when list of cols from the same table is not an expanded * expression

schemas/util.sql: added view_is_subset(view_def text)

schemas/util.sql: added view_is_automatically_updatable(view_def text)

bugfix: schemas/util.sql: show_create_view(): use the overridden version of pg_get_viewdef(), which supports expanded * expressions. this was possibly being used already whenever util happened to be in the search_path.

schemas/util.sql: lat_long_in_new_world(): use function rather than operator+search_path to allow inlining, which enables util.new_world() to only be evaluated once

bugfix: schemas/util.sql: operator @(postgis.geography, postgis.geography): must use wrapper function because st_coveredby() needs postgis to be in the search_path

fix: schemas/util.sql: point(): hide benign "Coordinate values were coerced into range [-180 -90, 180 90] for GEOGRAPHY" notices

schemas/util.sql: lat_long_in_new_world(): documented that this includes false positives above and below the New World bounding box, as described in util.bounding_box()

schemas/util.sql: bounding_box(): documented that the geography type stores all edges as arcs of great circles, resulting in the latitude lines bulging outward from the true bounding box. this will create false positives above and below the bounding box.

schemas/util.sql: added lat_long_in_new_world()

schemas/util.sql: added operator @(postgis.geography, postgis.geography). can't use && for this because it only compares 2D bounding boxes (which are geometry objects that do not support geocoordinate wraparound).

schemas/util.sql: added point()

schemas/util.sql: new_world(): removed no longer needed cast to postgis.geography

bugfix: schemas/util.sql: bounding_box(): must use postgis.geography (instead of postgis.geometry) because that handles geocoordinate wraparound correctly

bugfix: schemas/util.sql: bounding_box(): need to explicitly set SRID to make sure the correct value is used

bugfix: schemas/util.sql: bounding_box(): use st_makeenvelope() instead of st_makebox2d() because st_makebox2d() doesn't support geocoordinate wraparound (it is not SRID-aware)

schemas/util.sql: new_world(): removed no longer needed cast to postgis.geometry

schemas/util.sql: bounding_box(): return postgis.geometry instead of postgis.box2d because box2d is not directly used in postgis functions

schemas/util.sql: added new_world()

bugfix: schemas/util.sql: bounding_box(): use util.range instead of numrange to support wraparound ranges for geocoordinates

schemas/util.sql: range(numeric, numeric): use util.range instead of numrange to support wraparound ranges (which use a modulus system such as geocoordinates)

schemas/util.sql: added `range` type (and cast to numrange), which allows wraparound ranges such as for geocoordinates

schemas/util.sql: bounding_box(): documented usage

schemas/util.sql: added bounding_box()

schemas/util.sql: range(numeric, numeric): use simpler numrange() constructor instead of text input syntax

schemas/util.sql: operator functions: renamed to plain-text names to follow the Postgres convention of making the function name be the human-readable name of the operator

schemas/util.sql: added ~ operator for ranges

fix: schemas/util.sql: view_def_to_orig(): also need to merge .* expressions resulting from a SELECT * of a join, to avoid duplicated columns

fix: schemas/util.sql: view_def_to_orig(): require at least 6 cols to avoid false positives in the expansion match pattern

bugfix: schemas/util.sql: view_def_to_orig(): 1st col: can't prepend \y because it considers only \w chars, not "

bugfix: schemas/util.sql: view_def_to_orig(): don't match whitespace in the middle of a "" identifier, as this could throw off the parser

schemas/util.sql: eval(): restore user's intent by running util.view_def_to_orig() on the query to unexpand expanded * expressions

schemas/util.sql: pg_get_viewdef(): use util.view_def_to_orig()

schemas/util.sql: added view_def_to_orig()

fix: schemas/util.sql: pg_get_viewdef(): should be STABLE because it references system catalogs

schemas/util.sql: added pg_get_viewdef() wrapper, which unexpands expanded * expressions

bugfix: schemas/util.sql: mk_drop_from_create(): need to match first rather than last CREATE

schemas/util.sql: recreate_view(): support omitting the view_query if the view has already been modified (eg. for public.*_view, which allow changing the view as a separate step)

bugfix: schemas/util.sql: recreate(): need to handle case where util.mk_drop_from_create() is NULL

bugfix: schemas/util.sql: mk_drop_from_create(): only match CREATE if no custom DROP came before it

fix: schemas/util.sql: explain2notice_msg_if_can(): also need to catch invalid_cursor_definition ("cannot open multi-query plan as cursor")

schemas/util.sql: added copy()