Mutations are regular-expression-based replacements that are applied
after variants have been computed. They are meant to be used for
variations on character level.
Add spelling variations for German umlauts.
Only one addr: tag can be processed currently, so make
sure it is the one without suffixes to not get odd data.
addr:street is the exception because it uses a different
matching mechanism.
Using partial names turned out to not work well because there are
often similarly named streets next to each other. It also
prevents us from being able to take into account all addr:street:*
tags.
This change gets all the full term tokens for the addr:street tags
from the DB. As they are used for matching only, we can assume that
the term must already be there or there will be no match. This
avoid creating unused full name tags.
The highway key is being used more and more for non-ways these
days. This clashes with Nominatim's assumption that essentially
everything that has a highway tag can be used as the street part
of the address.
Change the default rank of highway objects to 30 to avoid this.
Only the known values for streets keep the rank 26 and are now
listed explicitly.
Check if the API script exists on the expected location before
running php-cli. This way we can add a useful hint about the
project directory.
Fixes#2513.
This mode gets updates until the server reports no new diffs
anymore.
Also adds additional indexing, when the main indexing step left
a couple of objects to process. This happens only when the
next update is expected to be more than 40min away.
Point-in-polygon queries are much faster with a SP-GIST geometry
index, so use that for the index used to check if a housenumber
is inside a building.
Only available with Postgis 3. There is an automatic fallback to
GIST for Postgis 2.
Adds a tagger for names by language so that the analyzer of that
language is used. Thus variants are now only applied to names
in the specific language and only tag name tags, no longer to
reference-like tags.
Implements per-name choice of analyzer. If a non-default
analyzer is choosen, then the 'word' identifier is extended
with the name of the ana;yzer, so that we still have unique
items.
Adds a mandatory section 'analyzer' to the token-analysis entries
which define, which analyser to use. Currently there is exactly
one, generic, which implements the former ICUNameProcessor.
Adds parsing of multiple variant lists from the configuration.
Every entry except one must have a unique 'id' paramter to
distinguish the entries. The entry without id is considered
the default. Currently only the list without an id is used
for analysis.
Sanatizer functions allow to transform name and address tags before
they are handed to the tokenizer. Theses transformations are visible
only for the tokenizer and thus only have an influence on the
search terms and address match terms for a place.
Currently two sanitizers are implemented which are responsible for
splitting names with multiple values and removing bracket additions.
Both was previously hard-coded in the tokenizer.
There is no need for the additional layer of indirection that
the ICUNameProcessorRules class adds. The ICURuleLoader can
fill the database properties directly.
Adds class, type, country and rank to the exported information
and removes the rather odd hack for countries. Whether a place
represents a country boundary can now be computed by the tokenizer.
When matching address parts from addr:* tags against place names,
the address names where so far converted to full names and compared
those to the place names. This can become problematic with the new
ICU tokenizer once we introduce creation of different variants
depending on the place name context. It wouldn't be clear which
variant to produce to get a match, so we would have to create all of
them. To work around this issue, switch to using the partial terms
for matching. This introduces a larger fuzziness between matches but
that shouldn't be a problem because matching is always geographically
restricted.
The search terms created for address parts have a different problem:
they are already created before we even know if they are going to be
used. This can lead to spurious entries in the word table, which slows
down searching. This problem can also be circumvented by using only
partial terms for the search terms. In terms of searching that means
that the address terms would not get the full-word boost, but given
that the case where an address part does not exist as an OSM object
should be the exception, this is likely acceptable.
A boolean check for dynamic changes of address parts is not
sufficient. The order of choice should be:
1. an addr:* part matches the name
2. the address part surrounds the object
3. the address part was declared as isaddress
The implementation uses a slightly different ordering
to avoid geometry checks unless strictly necessary (isaddress
is false and no matching address).
See #2446.
Adds a function to the Configuration class to load a YAML
file. This means that searching for the file is generalised
and works the same now for all configuration files. Changes
the search logic, so that it is always possible to have a
custom version of the configuration file in the project
directory.
Move ICU tokenizer to use new load function.
Linked places may bring in extra names. These names need to be
processed by the tokenizer. That means that the linking needs
to be done before the data is handed to the tokenizer. Move finding
the linked place into the preparation stage and update the name
fields. Everything else is still done in the indexing stage.
The new icu tokenizer is now no longer compatible with the old
legacy tokenizer in terms of data structures. Therefore there
is also no longer a need to refer to the legacy tokenizer in the
name.
This separates the logic of creating word sets from the Phrase
class. A tokenizer may now derived the word sets any way they
like. The SimpleWordList class provides a standard implementation
for splitting phrases on spaces.
Postgresql is very bad at creating statistics for jsonb
columns. The result is that the query planer tends to
use JIT for queries with a where over 'info' even when
there is an index.
The BDD tests cannot make assumptions about the structure of the
word table anymore because it depends on the tokenizer. Use more
abstract descriptions instead that ask for specific kinds of
tokens.
Requires a second wrapper class for the word table with the new
layout. This class is interface-compatible, so that later when
the ICU tokenizer becomes the default, all tests that depend on
behaviour of the default tokenizer can be switched to the other
wrapper.
Two replacement words directly following each other did not
work as expected because each expects a space at the
beginning/end while there was only one space available.
Also forbit composing a word after a space was added in the
end by a previous replacement.
The new format combines compound splitting and abbreviation.
It also allows to restrict rules to additional conditions
(like language or region). This latter ability is not used
yet.
Compound decomposition now creates a full name variant on
import just like abbreviations. This simplifies query time
normalization and opens a path for changing abbreviation
and compund decomposition lists for an existing database.
This adds precomputation of abbreviated terms for names and removes
abbreviation of terms in the query. Basic import works but still
needs some thorough testing as well as speed improvements during
import.
New dependency for python library datrie.
We've previously added searching through rank 30 in a house
number search to enable searches for house number+name.
This had the unintended side effect that rank 30 objects
are also returned in s search that dropped the house number
from the query. This is wrong because POIs cannot function
as a parent to a house number.
This fix drops all rank 30 objects from the results for a
house number search if they do not match the requested house
number.
Special terms need to be prefixed by a space because they are
full terms.
For countries avoid duplicate entries of word tokens.
Adds tests for adding country terms.
- only save partial words without internal spaces
- consider comma and semicolon a separator of full words
- consider parts before an opening bracket a full word
(but not the part after the bracket)
Fixes#244.
Explicitly check for the tokenizer source file to check that
the name is correct. We can't use the import error for that
because it hides other import errors like a missing
library.
Fixes#2327.
The ICU library only offers transliterations for a limited set of
script. Add transliterations for missing scripts from the PostgreSQL
module. These means that the same selection of scripts is supported
as with the old module.
The tokenizer to be used can be choosen with -DTOKENIZER.
Adapt all tests, so that they work with legacy_icu tokenizer.
Move lookup in word table to a function in the tokenizer.
Special phrases are temporarily imported from the wiki until
we have an implementation that can import from file. TIGER
tests do not work yet.
This adds an installation step for PHP code for the tokenizer. The
PHP code is split in two parts. The updateable code is found in
lib-php. The tokenizer installs an additional script in the
project directory which then includes the code from lib-php and
defines all settings that are static to the database. The website
code then always includes the PHP from the project directory.
The BDD tests still use the old-style amenity creation scripts
because we don't have simple means to import a hand-crafted
test file of special phrases right now.
The name analyzer is the actual work horse of the tokenizer. It
is instantiated on a thread-base and provides all functions for
analysing names and queries.
Indexing is now split into three parts: first a preparation step
that collects the necessary information from the database and
returns it to Python. In a second step the data is transformed
within Python as necessary and then returned to the database
through the usual UPDATE which now not only sets the indexed_status
but also other fields. The third step comprises the address
computation which is still done inside the update trigger in
the database.
The second processing step doesn't do anything useful yet.
Creating and populating the word table is now the responsibility
of the tokenizer.
The get_maxwordfreq() function has been replaced with a
simple template parameter to the SQL during function installation.
The number is taken from the parameter list in the database to
ensure that it is not changed after installation.
This adds the boilerplate for selecting configurable tokenizers.
A tokenizer can be chosen at import time and will then install
itself such that it is fixed for the given database import even
when the software itself is updated.
The legacy tokenizer implements Nominatim's traditional algorithms.
These tables have never been actively maintained and the code is
completely untested. With the upcomming changes, it is unlikely
that the code remains usable.
This removes the aux tables and all code that references them.
Rename function to reflect that it is only used for precomputation.
The token IDs are not really needed, so don't bother to compute
the array of tokens.
- the command is now --import-special-phrases
- the output is not an sql file anymore, data are directly imported to the database.
- the little part on the documentation (section data import) has been modified.
Drops all calls to PHP utility functions. nominatim cli functions
are used where possible, to stay as close to the final code as
possible with the tests.
By removing the PHP calls, the test code now only uses osm2pgsql and
the database module from the build directory.
Always look up the closest housenumber before looking up
interpolations. This ensures that closer housenumbers are
preferred over interpolations.
Fixes#2214.
Also switches to jinja-based preprocessing, which allows to
simplify the SQL files. Use 'if not exists' where possible
so that the step can be rerun to fix missing indexes.
Replaces various hand-crafted replacements of varying format with
a single Jinja2 templating mechanism. Allows full access to
configuration if necessary.
Instead of parsing the DSN for each external libpq program we
are going to execute, provide a function that feeds them all
necessary parameters through the environment.
osm2pgsql is the first user.
Psycopg2 has changed the kind of exception that is emitted on
deadlocks between versions 2.7 and 2.8. The code was already
trying to catch both kind of errors but because the
psycopg2.errors package is unknown in 2.7 and below, the
code would throw an exception on anything but a deadlock error.
This commit wraps the deadlock handling into a context manager
to avoid code duplication and uses module imports to detect if
the new error codes are available.
Also sets the required psycopg2 version to 2.7 or bigger as
versions below are difficult to test.
So far the data directory constant has pointed to the source
directory to be usable with different subdirectories. Now only
the data subdirectory itself is being used with the constant,
so point to the directory directly.
This replaces {data_dir}/settings throughout the code, so that
the configuration may be placed somewhere else in the directory
structure (e.g. in /etc).
With the website directory now tied to the project directory instead
of the build directory, it is no longer possible to use make for
running the web server.
Given that the module is now copied to the project directory
when no module path is set, we need the information that the
module path is empty. Therefore hand in the default module path
in a separate variable.
This is a exception to be thrown when the error occures because
of bad user data. We don't want to print a full stack trace in
these cases but just tell the user what went wrong.
The new functions always creates normal and partitioned functions.
Also adds specialised connection and cursor classes for adding
frequently used helper functions.
Given that only one command will be executed in the end, it is
not necessary to import what amounts to the whole library. This
becomes in particular important for update functions that have
a dependency on pyosmium. The dependency can remain optional for
people not using updates.
The PHP scripts need to know the position of the nominatim
tool in order to call it. This is handed in as environment
variable, so it can be set by the Python script.
In the future, the BDD tests will simply set up the required
test database themselves. Like with the template database, it
is not reimported when it already exists unless that is explicitly
forced.
Makes most of the API tests currently fail because they still
point to old test data.
Introduces a new class DBRow that encapsulates the comparison
functions. This also is responsible for formatting more informative
assert messages. place and placex steps are unified.
Put the connection to the test database into auto-commit mode
and get rid of the explicit commits. Also use cursors always in
context managers and unify the two implementations that copy
data from the place table.
The project directory contains the website script as
configured through the test configuration. This means
that tests are now completely independet of any
configuration that may be contained in the build
directory.
Also removes the hack to inject additional settings via
a environment variable.
DB tests now can simply set the environment to change configuration
variables. API tests still rely on a configuration file.
Also, query.php needs to set up the CONST_* variables to work with
the query scripts. That is a tiny bit messy and duplicates code
but this part will need to be reworked later.
CONST_BasePath is split into separate configuration variables
for binaries, libraries and data. These variables as well as
the installation path are now set in the executable directly and
no longer configurable via project settings.
This is the first step towards an installable software. The
executables should know per installation where to find their
necessary data to execute. Project configuration needs to be
restricted to settings that really concern the specific Nominatim
installation.
Update names in the coutry_names table on the fly from incomming
OSM country data. Adding a small sanity check that the country
must be an OSM relation and within the area where we expect the
country to be.
If a place node is already linked against a boundary, it should not
be used for linking again. It is usually a sign of a mapping error,
when there are multiple boundary candidates. This change just avoids
inconsistent data in the database, it does not guarantee that the
linking is against the more correct boundary.
Rank 30 objects usually use the address parts of their parent.
When the parent has address parts that are areas but not marked
as isaddress, then the parent might go through multiple administrative
areas. In that case recheck if the right area has been choosen
for the object in question instead of relying on isaddress.
Note that we really only have to do the recomputation in the
case of 'isarea = True and isaddress = False' which hopefully
keeps the number of additional geometric operations we have to do
to a minimum.
There is one more special case to be taken into account here: a
street may go through two administrative areas and a house along
that street is placed in one of the area while the addr:* tags
says it belongs to the other. In that case we must not switch
the isaddress to the one it is situated. To avoid that recheck
the address names against the name of the ara. That is not perfect
but should cover most cases.
Fixes#328.
Multi-word partial terms had an undue advantage over separate partial
terms because they only need to pay the penalty once. This changes
the behaviour by setting the penalty according to the number of
words in the token. This should get rid of search interpretations
with low chance of matching.
This also fixes handling of exact term matching. We now match against
all exact terms of the query, not just a couple of them collected
while building the interpretations.
Also adds a penalty to very short postcodes.
House numbers need special handling because they may appear after
the street term. That means we canot just use them as the main name
for searches where the address has its own search term entries.
Doing this right now, we are able to find '40, Main St, Town' but not
'Main St 40, Town'.
This switches to using the housenumber token as the name term instead.
House number tokens can get special handling when building the search
query that covers the case where they come after the street.
The main disadvantage is that this once more increases the numbers
of possible search interpretation of which we have already too many.
no penalty for housenumber searches
The previous behaviour was a left-over from a former version
where such POIs parented to the street. Now that they parent to
places, it should be included.
get_addressdata() now also checks if the place itself has entries
in the place_addressline table and merges them into the results.
Also restrict checking for address tag places to cases where the
name cannot be found in the parent's address search terms. Looking
up all address tags is just too slow.
While previously the content of addr:* tags was only added
to the list of address search keywords, we now really look up
the matching place. This has the advantage that we pull in all
potential translations from the place, just like all the other
address terms that are looked up by neighbourhood search.
If no place can be found for a given name, the content of the
addr:* tag is still added to the search keywords as before.
Go back to using centroid when determining if one admin level
is within another. There are cases where boundaries are slightly
misaligned due to mapping errors (not using the same ways in the
relations).
Only declare boundaries the same when they have the same wikidata
tag _and_ have exactly the same geometry. This works around tagging
errors with the wikidata tag, which happen because of automated
edits to the wikidata tag.
The Polish community maps admin boundaries that span multiple
levels by duplicating the boundary relations. Detect this situation
by looking out for matching wikidata tags. The higher ranked
duplicates are then thrown out from the address pool by setting
their address rank to 0.
When a POI has no addr:street but an addr:place that is not
contained in the name list of the parent place, then remember
this situation and merge the content of addr:place into the
address output.
We don't need to care about translations in this case because
it is obvious that no object with translations exists if the
parent isn't the object named in addr:place.
If an addr:place is given but no addr:street tag, then bind
the rank 30 object always to a <=25 object, even when there
is none found with the same name.
When a place of rank 30 has addr tags that are not covered by the
search terms of the parent, add a separate entry for the POI in
the search_name table that includes the addr tags. We can only
do that with named places. For POIs without a name the housenumber
is used as name. If that is not available either, searching still
won't work.
place=postcode places are artificial places that collect addr:postcode
points for aggration. They should neither show up in the address nor
be searchable. That means that there is no need to index them at all.
Only let boundary=postal_code through which define correct areas for
postcodes.
Boundaries shound derive the address part type from the
linked place if possible. This was already implemented
for the address objects but not for the address information
from the address itself.
Fixes#1949.
It may happen that two different postcodes normalize to exactly
the same token. In that case we still need two different entries
in the word table. Token lookup will then make sure that the correct
one is choosen.
Fixes#1953.
Add a section on setting up the development environment which now
also includes the former chapter on recreating the documentation.
Move the README from test/ into the manual as the new Testing
chapter.
Squares are now addressable (on address level 25) and thus can
be attached to a house number via addr:place. Needed to increase
the rank range for matching up addr:place to 25.
Always add contents of addr:* tags into address part of the search
table, even when there is no corresponding other name. This keeps
search tolerant to the kind of tagging where parts show up in the
address that have no corresponding object in the database or where
it is only an unaddressable object.
Before updating an admin boundary we need to make sure that any
artificially generated 'linked_place' entry is removed from the
extratags column. This ensures that the place designation does
not linger when a linked place disappears and that it is updated
when the linking changes.
