pyrightconfig.json :
for the paths searx, searxng_extra and tests, individual rules were
defined (for example, in test fewer / different rules are needed than in the
searx package
searx/engines/__builtins__.pyi :
The builtin types that are added to the global namespace of a module by the
intended monkey patching of the engine modules / replaces the previous
filtering of the stdout using grep.
test.pyright_modified (utils/lib_sxng_test.sh) :
static type check of local modified files not yet commited
make test :
prerequisite 'test.pyright' has been replaced by 'test.pyright_modified'
searx/engines/__init__.py, searx/enginelib/__init__.py :
First, minimal typifications that were considered necessary.
The HTTP X-Forwarded-Proto (XFP) request header is a *de-facto* standard header
for identifying the protocol (HTTP or HTTPS) that a client used to connect to a
proxy or load balancer.[1]
The ``X-Scheme`` header was added 10 years ago, why ``X-Scheme`` was used back
then and not ``X-Forwarded-Proto``, nobody knows today / possibly because
``X-Forwarded-Proto`` wasn't a *de-facto* standard back then.
[1] https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/Headers/X-Forwarded-Proto
[2] https://github.com/searx/searx/commit/6ef7c3276
* [fix] CI task "update_engine_traits.py" fails
To catch all problems with an HTTP request, the more general class
``httpx.HTTPError`` must be caught, for your test use::
$ ./manage dev.env
$ python ./searxng_extra/update/update_engine_traits.py
Closes: https://github.com/searxng/searxng/issues/5068
* [data] update searx.data - update_engine_traits.py
The error message in case the vqd value could not be determined was incorrect
and triggered an exception::
File "/usr/local/searxng/searxng-src/searx/engines/duckduckgo.py", line 132, in get_vqd
logger.error("vqd value from duckduckgo.com ", resp.status_code)
Message: 'vqd value from duckduckgo.com '
Arguments: (202,)
If the ``on_result`` handle returns False, then the ``else`` was always jumped
to, which throws the NotImplementedError exception::
File "/usr/local/searxng/searxng-src/searx/results.py", line 99, in extend
raise NotImplementedError(f"no handler implemented to process the result of type {result}")
NotImplementedError: no handler implemented to process the result of type MainResult(title=...
Pyright has been installed twice so far, once via `package.json` and once in the
test script via `npx --no-install`. Separating the type checks in the CI and on
the developer desktop is also not necessary.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Implement rules and functions to format shell scripts:
$ make format.shell
or alternatively to format all source code:
$ make format
The formatter `shfmt` reads the rules from the editorconfig[1]
If any EditorConfig files are found, they will be used to apply formatting
options. If any parser or printer flags are given to the tool, no
EditorConfig files will be used.
[1] https://github.com/patrickvane/shfmt?tab=readme-ov-file#description
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Whether the query is a real calculation tasks is currently only detected in the
AST, resulting in unnecessary creatins of subprocesses. This problem is
mitigated with this patch: if the query contains letters, it is obviously not a
math problem, and the plugin can return without further action.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
In customizing it should be decided which plugin modules should be loaded and
which should not.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Depending on the respective runtime behavior, it could happen that the initial
loading of the DB tables in the cache was performed multiple times and in
parallel. The concurrent accesses then led to the `sqlite3.OperationalError:
database is locked` exception as in #4951.
Since this problem depends significantly on the runtimes (e.g., how long it
takes to retrieve the content for a table), this error could not be observed in
all installations.
Closes: https://github.com/searxng/searxng/issues/4951
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
The issue was introduced in commit: edfbf1e
Problematic code::
def timeout_func(timeout, func, *args, **kwargs):
...
if not p.is_alive():
ret_val = que.get()
else:
logger.debug("terminate function after timeout is exceeded") # type: ignore
p.terminate()
p.join()
p.close()
The `logger` function in the `else` path is not defined. Was accidentally
removed in commit edfbf1e without providing an appropriate replacement.::
File "/usr/local/searxng/searx/plugins/calculator.py", line 216, in timeout_func
logger.debug("terminate function after timeout is exceeded") # type: ignore
^^^^^^
NameError: name 'logger' is not defined
The exception triggered by this prevents the `p.terminate()` from being
executed. As a result, the processes accumulate in memory (memory leak).
Related: https://github.com/searxng/searx-instances/discussions/708#discussioncomment-13688168
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
On demand, the tracker data is loaded directly into the cache, so that the
maintenance of this data via PRs is no longer necessary.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
With the new implementation for removing tracker arguments from URLs, we use the
tracker definitions of the ClearURLs project [1]. These definitions are
regularly updated and significantly improve data protection, which is why we
should now also enable this plugin by default.
[1] https://github.com/ClearURLs
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
The code formatter allows up to 120 characters per line, but for better
readability the editor should prefer 80 characters per line
In the HTML templates (jinja) we usually use an indentation of two and no tabs.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
checker.yml
1. The checker is not yet of sufficient quality to allow the results of the
check to be evaluated / we do not evaluate them ourselves.
2. The checker sends hundreds of requests to the search engines and causes
problems there / we either overload small providers or we train their bot
defenses to use the SearXNG signature.
documentation.yml
Building the documentation and deploying it on GH-docs of a clones (GH forks) is
generally not desirable either --> We have >2k clones, but we only need one
up-to-date documentation and that is the one from the master branch of the
searxng/searxng repo.
If search engines like Google start linking to the documentation in the clones,
SearXNG users may no longer find the original documentation or be lost in the
flood of options.
Related:
- https://github.com/searxng/searxng/issues/4847
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
The types necessary for weather information such as GeoLocation, DateTime,
Temperature,Pressure, WindSpeed, RelativeHumidity, Compass (wind direction) and
symbols for the weather have been implemented.
There are unit conversions and translations for weather property labels.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
The previous implementation for determining the description of an engine did not
take into account that the UI languages can also have a region tag and/or a
script tag:
el-GR: Ελληνικά, Ελλάδα (Greek, Greece)
fa-IR: فارسی, ایران (Persian, Iran)
nb-NO: Norsk bokmål, Norge (Norwegian bokmål, Norway)
nl-BE: Nederlands, België (Dutch, Belgium)
pt-BR: Português, Brasil (Portuguese, Brazil)
zh-HK: 中文, 中國香港特別行政區 (Chinese, Hong Kong SAR China)
zh-Hans-CN: 中文, 中国 (Chinese, China)
zh-Hant-TW: 中文, 台灣 (Chinese, Taiwan)
Closes: https://github.com/searxng/searxng/issues/4842
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Sometimes (e.g. when ddg does not have a result item) there is no content and
the engine will fail with an IndexError:
* Error: IndexError
* Percentage: 10
* Parameters: `()`
* File name: `searx/engines/duckduckgo.py:375`
* Function: `response`
* Code: `item["content"] = extract_text(eval_xpath(div_result, './/a[contains(@class, "result__snippet")]')[0])`
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
There is currently no known z-library, and all known URLs are dead [1]. To avoid
interrupting automated updates, a connection error to a z-library is treated as
a *known error*, and the old properties of the z-library are retained.
[1] https://github.com/searxng/searxng/issues/3610
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
In the previous implementation, all databases were loaded into memory when
importing the searx.data package, regardless of whether they were ever needed.
Regardless of this, it is an antipattern to load entire databases into memory
when importing a package or module; databases should be loaded when needed.
Lazy loading is a first step toward improving memory usage and also improves
performance when setting up the runtime environment. Building on this,
subsequent PRs will be able to further optimize memory behavior, e.g., by using
a real database application such as the one already available via
searx.cache.ExpireCache
Related:
- https://github.com/searxng/searxng/discussions/1892
- https://github.com/searxng/searxng/pull/3458
- https://github.com/searxng/searxng/pull/4650
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
The global variable CACHE is not initialized when DDG images or DDG videos
import the get_vqd() function (please remember: the engine modules are imported
using the importlib method and not via the `import` keyword).
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Chapter on the purpose of (git) commits
The commits and their messages are elementary for the traceability of changes
and are unfortunately still too often given too little attention.
It therefore seems necessary to dedicate a chapter to this topic in the context
of development.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Starting with ionicons-8.0.8 the SVG already contains a class attribute and
instaed of using SVGO plugin ``addAttributesToSVGElement`` we habve to use
``addClassesToSVGElement`` to add out ``__jinja_class_placeholder__``.
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
The EngineCache class replaces all previously individual solutions for caches in
the context of the engines.
- demo_offline.py
- duckduckgo.py
- radio_browser.py
- soundcloud.py
- startpage.py
- wolframalpha_api.py
- wolframalpha_noapi.py
Search term to test most of the modified engines::
!ddg !rb !sc !sp !wa test
!ddg !rb !sc !sp !wa foo
For introspection of the DB, jump into developer environment and run command to
show cache state::
$ ./manage pyenv.cmd bash --norc --noprofile
(py3) python -m searx.enginelib cache state
cache tables and key/values
===========================
[demo_offline ] 2025-04-22 11:32:50 count --> (int) 4
[startpage ] 2025-04-22 12:32:30 SC_CODE --> (str) fSOBnhEMlDfE20
[duckduckgo ] 2025-04-22 12:32:31 4dff493e.... --> (str) 4-128634958369380006627592672385352473325
[duckduckgo ] 2025-04-22 12:40:06 3e2583e2.... --> (str) 4-263126175288871260472289814259666848451
[radio_browser ] 2025-04-23 11:33:08 servers --> (list) ['https://de2.api.radio-browser.info', ...]
[soundcloud ] 2025-04-29 11:40:06 guest_client_id --> (str) EjkRJG0BLNEZquRiPZYdNtJdyGtTuHdp
[wolframalpha ] 2025-04-22 12:40:06 code --> (str) 5aa79f86205ad26188e0e26e28fb7ae7
number of tables: 6
number of key/value pairs: 7
In the "cache tables and key/values" section, the table name (engine name) is at
first position on the second there is the calculated expire date and on the
third and fourth position the key/value is shown.
About duckduckgo: The *vqd coode* of ddg depends on the query term and therefore
the key is a hash value of the query term (to not to store the raw query term).
In the "properties of ENGINES_CACHE" section all properties of the SQLiteAppl /
ExpireCache and their last modification date are shown::
properties of ENGINES_CACHE
===========================
[last modified: 2025-04-22 11:32:27] DB_SCHEMA : 1
[last modified: 2025-04-22 11:32:27] LAST_MAINTENANCE :
[last modified: 2025-04-22 11:32:27] crypt_hash : ca612e3566fdfd7cf7efe2b1c9349f461158d07cb78a3750e5c5be686aa8ebdc
[last modified: 2025-04-22 11:32:30] CACHE-TABLE--demo_offline: demo_offline
[last modified: 2025-04-22 11:32:30] CACHE-TABLE--startpage: startpage
[last modified: 2025-04-22 11:32:31] CACHE-TABLE--duckduckgo: duckduckgo
[last modified: 2025-04-22 11:33:08] CACHE-TABLE--radio_browser: radio_browser
[last modified: 2025-04-22 11:40:06] CACHE-TABLE--soundcloud: soundcloud
[last modified: 2025-04-22 11:40:06] CACHE-TABLE--wolframalpha: wolframalpha
These properties provide information about the state of the ExpireCache and
control the behavior. For example, the maintenance intervals are controlled by
the last modification date of the LAST_MAINTENANCE property and the hash value
of the password can be used to detect whether the password has been changed (in
this case the DB entries can no longer be decrypted and the entire cache must be
discarded).
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
The Yahoo engine's fetch_traits function has been encountering an error in CI
jobs for several months [1], thus aborting the process for all other engines as
well.
The language selection dialog (which fetch_traits calls) requires an `EuConsent`
cookie. Strangely, the cookie is not needed for searching, which is why the
engine itself still works.
Since Yahoo won't be conquering any new marketplaces in the foreseeable future,
it should be sufficient to hard-implement the list of currently available
languages (`yahoo_languages`).
[1] https://github.com/searxng/searxng/actions/runs/14720458830/job/41313149268
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
BTW: fix a bug with sys.path: repo-root (not util) needs to added to generate
autodoc from scripts in ./searxng_extra
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
On a long-running server, the tracebacks below can be found (albeit rarely),
which indicate problems with NoneType where a string or another data type is
expected.
result.img_src::
File "/usr/local/searxng/searxng-src/searx/templates/simple/result_templates/images.html", line 13, in top-level template code
<img src="" data-src="{{ image_proxify(result.img_src) }}" alt="{{ result.title|striptags }}">{{- "" -}}
^
File "/usr/local/searxng/searxng-src/searx/webapp.py", line 284, in image_proxify
if url.startswith('//'):
^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'startswith'
result.content::
File "/usr/local/searxng/searxng-src/searx/result_types/_base.py", line 105, in _normalize_text_fields
result.content = WHITESPACE_REGEX.sub(" ", result.content).strip()
~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
TypeError: expected string or bytes-like object, got 'NoneType'
html_to_text, when html_str is a NoneType::
File "/usr/local/searxng/searxng-src/searx/engines/wikipedia.py", line 190, in response
title = utils.html_to_text(api_result.get('titles', {}).get('display') or api_result.get('title'))
File "/usr/local/searxng/searxng-src/searx/utils.py", line 158, in html_to_text
html_str = html_str.replace('\n', ' ').replace('\r', ' ')
^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'replace'
presearch engine, when json_resp is a NoneType::
File "/usr/local/searxng/searxng-src/searx/engines/presearch.py", line 221, in response
results = parse_search_query(json_resp.get('results'))
File "/usr/local/searxng/searxng-src/searx/engines/presearch.py", line 161, in parse_search_query
for item in json_results.get('specialSections', {}).get('topStoriesCompact', {}).get('data', []):
^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'get'
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
This patch brings two major changes:
- ``Result.filter_urls(..)`` to pass a filter function for URL fields
- The ``enabled_plugins:`` section in SearXNG's settings do no longer exists.
To understand plugin development compile documentation:
$ make docs.clean docs.live
and read http://0.0.0.0:8000/dev/plugins/development.html
There is no longer a distinction between built-in and external plugin, all
plugins are registered via the settings in the ``plugins:`` section.
In SearXNG, plugins can be registered via a fully qualified class name. A
configuration (`PluginCfg`) can be transferred to the plugin, e.g. to activate
it by default / *opt-in* or *opt-out* from user's point of view.
built-in plugins
================
The built-in plugins are all located in the namespace `searx.plugins`.
.. code:: yaml
plugins:
searx.plugins.calculator.SXNGPlugin:
active: true
searx.plugins.hash_plugin.SXNGPlugin:
active: true
searx.plugins.self_info.SXNGPlugin:
active: true
searx.plugins.tracker_url_remover.SXNGPlugin:
active: true
searx.plugins.unit_converter.SXNGPlugin:
active: true
searx.plugins.ahmia_filter.SXNGPlugin:
active: true
searx.plugins.hostnames.SXNGPlugin:
active: true
searx.plugins.oa_doi_rewrite.SXNGPlugin:
active: false
searx.plugins.tor_check.SXNGPlugin:
active: false
external plugins
================
SearXNG supports *external plugins* / there is no need to install one, SearXNG
runs out of the box.
- Only show green hosted results: https://github.com/return42/tgwf-searx-plugins/
To get a developer installation in a SearXNG developer environment:
.. code:: sh
$ git clone git@github.com:return42/tgwf-searx-plugins.git
$ ./manage pyenv.cmd python -m \
pip install -e tgwf-searx-plugins
To register the plugin in SearXNG add ``only_show_green_results.SXNGPlugin`` to
the ``plugins:``:
.. code:: yaml
plugins:
# ...
only_show_green_results.SXNGPlugin:
active: false
Result.filter_urls(..)
======================
The ``Result.filter_urls(..)`` can be used to filter and/or modify URL fields.
In the following example, the filter function ``my_url_filter``:
.. code:: python
def my_url_filter(result, field_name, url_src) -> bool | str:
if "google" in url_src:
return False # remove URL field from result
if "facebook" in url_src:
new_url = url_src.replace("facebook", "fb-dummy")
return new_url # return modified URL
return True # leave URL in field unchanged
is applied to all URL fields in the :py:obj:`Plugin.on_result` hook:
.. code:: python
class MyUrlFilter(Plugin):
...
def on_result(self, request, search, result) -> bool:
result.filter_urls(my_url_filter)
return True
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
overflow-y: scroll
If the information on the image exceeds the visible area in the detail view,
there must be a way to access this information via scroll.
max-height: -17rem versus 7rem
There are new fields in the lower area, so the height of the image (top) must
be reduced
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Fix the issues reported by sphinx build::
docstring of searx.engines.google.max_page:1: ERROR: Unknown target name: "google: max 50 pages".
docstring of searx.engines.google_images.max_page:1: ERROR: Unknown target name: "google: max 50 pages".
docstring of searx.engines.google_scholar.max_page:1: ERROR: Unknown target name: "google: max 50 pages".
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
Markus Heiser
