Error reporting currently takes place for each error. This can result
in many error messages if a server has connectivity issues (i.e. when
it re-connects to the internet every 24 hours).
This commit adds a new option to the configuration file to define the
number of error reports to suppress before returning an error message
to the user.
Error reports are cached and therefore automatically purged after 24
hours. A successful bridge request does **not** clear the error count
as sporadic issues can be the result of actual problems on the server.
The implementation currently makes no assumption on the type of error,
which means it also suppresses bridge errors in debug mode. The default
value is, however, set to 1 which means all errors are reported.
References #994
Bridge errors are currently included as part of the feed to
notify users about erroneous bridges (before that, bridges
silently failed).
This solution, however, can produce a high load of error
messages if servers are down (see #994 for more details).
Admins may also not want to include error messages in feeds
in order to keep those kind of problems away from users or
simply to silently fail by choice.
This commit adds a new configuration section "error" with
one option "output" which can be set to following values:
"feed": To include error messages in the feed (default)
"http": To return a HTTP header for each error
"none": To disable error reporting
Note that errors are always logged to 'error.log' independent
of the settings above.
Closes#1066
* [ParameterValidator] Ensure context has all fields
Previously if a bridge had a set of parameters like:
const PARAMETERS = array(
'ContextA' => array(
'Param1' => array(
'name' => 'Param1',
'required' => true
)
),
'ContextB' => array(
'Param1' => array(
'name' => 'Param1',
'required' => true
),
'Param2' => array(
'name' => 'Param2',
'required' => true
)
)
)
and a query specifying both Param1 and Param2 was provided a 'Mixed
context parameters' error would be returned. This change ensures
ContextA in the above example would not be considered a relevant context.
RSS-Bridge currently has to guess the queried context from the data
provided by the user. This, however, can cause issues for bridges
that have multiple contexts with conflicting parameters (i.e. none).
This commit adds context hinting to queries via '&context=<context>'
which can be omitted in which case the context is determined as before.
The format factory can be based on the abstract factory class if it
wasn't static. This allows for higher abstraction and makes future
extensions possible. Also, not all parts of RSS-Bridge need to work
on the same instance of the factory.
References #1001
The cache factory can be based on the abstract factory class if it
wasn't static. This allows for higher abstraction and makes future
extensions possible. Also, not all parts of RSS-Bridge need to work
on the same instance of the factory.
References #1001
The bridge factory can be based on the abstract factory class if it
wasn't static. This allows for higher abstraction and makes future
extensions possible. Also, not all parts of RSS-Bridge need to work
on the same instance of the bridge factory.
References #1001
RSS-Bridge currently sanitizes the format name only for the display
action, which can cause problems if other actions depend on formats
as well.
It is therefore better to do sanitization in the factory class for
formats. Additionally, formats should not require a perfect match,
so 'Atom' and 'aToM' make no difference. This will also allow users
to define formats in their own style (i.e. only lowercase via CLI).
References #1001
Response headers may contain fields with no values.
Example:
"Referrer-Policy: "
In this case the current implementation of explode() results in an
error because there is no content after ": ". Changing the delimiter
to ":" and trimming the value manually fixes that issue.
Users currently only get one option: to open a new issue on GitHub.
This can, however, result in duplicate issues, which is not desired.
This commit adds a second button to the error message, which links
to the GitHub issues tracker with the search query set to find
errors for the current bridge. That way, users can collaborate
on the same issue.
Incorrect configuration values are currently handled individually
for each condition, resulting in a lot of repetitive operations.
This commit adds two new private functions to report errors to the
user and end execution of the script.
The configuration files are currently hard-coded in the configuration
classes and error messages. However, the implementation should not
rely on specific details like the file name. Instead, the files should
be part of the global definition.
This commit introduces two global constants for the configuration files
- FILE_CONFIG => 'config.ini.php'
- FILE_CONFIG_DEFAULT => 'config.default.ini.php'
RSS-Bridge currently statically sets the timezone to UTC which can
result in incorrect timestamps if the server is hosted in another
region.
This commit adds a new configuration parameter to allow admins to
specify their own timezone for their servers. Invalid values will
result in an error message.
Example:
[system]
timezone = "UTC"
For compatibility reasons the default value is set to UTC.
This parameter accepts any of the supported timezones listed at
https://www.php.net/manual/en/timezones.phpCloses#956
References #1001
If the bridge name matches exactly, it is not necessary to perform
a strtolower compare of bridges. In some situations this can lead
to much faster response times (depending on the amount of bridges
in whitelist.txt).
Default bridges are currently statically defined in index.php, which
is not the right place if we want to keep responsibilities separated.
This commit introduces a new file whitelist.default.txt that holds
the default bridges and which is loaded automatically, if whitelist.txt
doesn't exist.
Due to this it is also no longer necessary to have write permission
for the root directory.
References #1001
This reverts commit 052844f5e1.
There is a bug in ->remove() that causes the parser to incorrectly
identify elements in the DOM tree that shouldn't exist anymore.
References #1151
simplehtmldom 1.9 introduced new functions to recursively remove
nodes from the DOM. This allows removing elements without the need
to re-load the document by using $html->load($html->save()), which
is very inefficient.
Find more information about remove() at
https://simplehtmldom.sourceforge.io/docs/1.9/api/simple_html_dom_node/remove/
find('*') wasn't supported in older versions of simplehtmldom but it
is supported now. Thus, all custom implementations can be replaced
by the correct solution.
This fixes the following issue:
1. bridge sets unique ids for the items (ids get hashed)
2. items go to the cache
3. on next run items get loaded from cache
4. these items have different ids because they were hashed again
5. they show up twice in feed reader
- For consistency, functions should always return null on non-existing data.
- WordPressPluginUpdateBridge appears to have used its own cache instance in the past. Obviously not used anymore.
- Since $key can be anything, the cache implementation must ensure to assign the related data reliably; most commonly by serializing and hashing the key in an appropriate way.
- Even though the default path for storage is perfectly fine, some people may want to use a different location. This is an example how a cache implementation is responsible for its requirements.
This commit adds support for a new parameter which specifies the type
of cache to use for caching. It is specified in config.ini.php:
[cache]
type = "..."
Currently only one type of cache is supported (see /caches). All uses
of 'FileCache' were replaced by this configuration option.
Note: Caching currently depends on files and folders (due to FileCache).
Experience may vary depending on the selected cache type. For now always
check if FileCache is working before testing alternative types.
References #1000
'uid' represents the unique id for a feed item. This item is null by
default and can be set to any string value. The provided string value
is always hashed to sha1 to make it the same length in all cases.
References #977, #1005
Add transformation from legacy items to FeedItems, before transforming
items to the desired format. This allows using legacy bridges alongside
bridges that return FeedItems.
As discussed in #940, instead of throwing exceptions on invalid
parameters, add messages to the debug log instead
Add support for strings to setTimestamp(). If the provided timestamp
is a string, automatically try to parse it using strtotime().
This allows bridges to simply use `$item['timestamp'] = $timestamp;`
instead of `$item['timestamp'] = strtotime($timestamp);`
Support simple_html_dom_node as input paramter for setURI
Support simple_html_dom_node as input parameter for setContent
* core: Add bridge parameter auto detection
This adds a new 'detect' action which accepts a URL from which an
appropriate bridge is selected and relevant parameters are extracted.
The user is then automatically redirected to the selected bridge.
For example to get a feed from: https://twitter.com/search?q=%23rss-bridge
we could send a request to:
'/?action=detect&format=Atom&url=twitter.com/search%3Fq%3D%2523rss-bridge'
which would redirect to:
'/?action=display&q=%23rss-bridge&bridge=Twitter&format=Atom'.
This auto detection happens on a per-bridge basis, so a new function
'detectParameters' is added to BridgeInterface which bridges may implement.
It takes a URL for an argument and returns a list of parameters that were
extracted, or null if the URL isn't relevant for the bridge.
* [TwitterBridge] Add parameter auto detection
* [BridgeAbstract] Add generic parameter detection
This adds generic "paramater detection" for bridges that don't have any
parameters defined. If the queried URL matches the URI defined in the
bridge (ignoring https://, www. and trailing /) an emtpy list of parameters is
returned.
This commit adds a cache for 'getContents' to '/cache/server'. All
contents are cached by default (even in debug mode). If debug mode
is enabled, the cached data is overwritten on each request.
In normal mode RSS-Bridge adds the 'If-Modified-Since' header with
the timestamp from the previously cached data (if available) to the
request.
Find more information on 'If-Modified-Since' here:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/If-Modified-Since
If the server responds with "304 Not Modified", the cached data is
returned.
If the server responds with "200 OK", the received data is written
to the cache (creates a new cache file if it doesn't exist yet).
No changes were made for all other response codes.
Servers that don't support the 'If-Modified-Since' header, will
respond with "200 OK".
For servers that respond with "304 Not Modified", the required band-
width will decrease and RSS-Bridge will responding faster.
Files in the cache are forcefully removed after 24 hours.
Notice: Only few servers actually do support 'If-Modified-Since'.
Thus, most bridges won't be affected by this change.
simple_html_dom currently doesnt support "->find('*')", which is a
known issue: https://sourceforge.net/p/simplehtmldom/bugs/157/
The solution implemented by RSS-Bridge is to find all nodes WITHOUT
a specific attribute. If the attribute is very unlikely to appear
in the DOM, this is essentially returning all nodes.
This is the meaning behind
"->find('*[!b38fd2b1fe7f4747d6b1c1254ccd055e]')"
Error log reports "PHP Notice: Undefined offset: 2 in /rss-bridge/
lib/Debug.php on line 112" if the array returned by debug_backtrace
does not contain 3 items.
This commit fixes the issue by always using the last element in the
backtrace "end($backtrace)".
- Initialize with null to prevent leaking configurations
- Check if the working directory is a directory
- Store the real path instead of raw data
- Add final path separator as expected by Bridge::create
general
- Use self:: instead of Cache:: or static::
- Rename $dirCache to $workingDir
- Initialize $workingDir with null
function setDir
- Clear previous working directory before checking input parameters
- Change wording for the exception messages
- Store realpath instead of raw parameter
- Add path separator
This ensures the path always ends with the path separator, as assumed
by Cache::create
- Add check if the provided working directory is a valid directory
function getDir
- Use static parameter instead of function variable
- Change wording for the exception message
function create
- Rename parameter $nameCache to $name
- Rename $pathCache to $filePath
- Change wording for the exception messages
function isValidNameCache
- Rename function to isCacheName
- Rename parameter $nameCache to $name
- Explain in the function documentation the meaning of a 'valid name'
- Ensure Boolean return value (preg_match returns integer)
- Check if $name is a string
- Use slashes to enclose the regex
This is the first step in adding documentation to the core library
of RSS-Bridge. The documentation is not yet extracted by phpdoc,
yet may prove useful to anyone interested in starting with RSS-Bridge.
Commit e26d61e introduced a bug that causes the error message "The
bridge you [sic!] looking for does not exist." if the bridge name
specified in the query ends on "Bridge"
(i.e. '&bridge=SoundcloudBridge'), while other queries work fine
(i.e. '&bridge=Soundcloud').
This commit fixes that issue by sanitizing the bridge name before
creating the class.
References #922
- Move all whitelisting functionality inside Bridge.php
- Set default whitelist once in index.php using Bridge::setWhitelist()
- Include bridge sanitizing inside Bridge.php
Bridge::sanitizeBridgeName($name)
Bridge.php now maintains the whitelist internally.
Also adds documentation to Debug.php!
* Debug::isEnabled()
Checks if the DEBUG file exists on disk on the first call (stored in
memory for the duration of the instance). Returns true if debug mode
is enabled for the client.
This function also sets the internal flag for Debug::isSecure()!
* Debug::isSecure()
Returns true if debuging is enabled for specific IP addresses, false
otherwise. This is checked on the first call of Debug::isEnabled().
If you call this function before Debug::isEnabled(), the default value
is false.
Replaces 'debugMessage' by specialized debug function 'Debug::log'.
This function takes the same arguments as the previous 'debugMessage'.
A separate Debug class allows for further optimization and separation
of concern.
- PATH_LIB_BRIDGES defines the path to bridges
- PATH_LIB_FORMATS defines the path to formats
- PATH_LIB_CACHES defines the path to caches
Include constants in RssBridge.php for consistency
Move CACHE_DIR from index.php to /lib/RssBridge.php and change name
to PATH_CACHE.
PATH_CACHE is one of the core paths of RSS-Bridge and should therefore
be defined in the core file RssBridge.php.
- Remove file documentation and license remark (defined in repository
scope - see README / UNLICENSE)
- Remove example usage (if necessary should be included in the Wiki)
"The require_once statement is identical to require except PHP will
check if the file has already been included, and if so, not include
(require) it again."
-- http://php.net/manual/en/function.require-once.php
Vendor files (simple_html_dom.php and urljoin.php) are included in the
repository and therefore shipped with all releases. If one of the files
is missing, either the repository or the release is incomplete.
PHP will generate error messages if either of the files is missing, so
there is no need to check availability manually unless it is done for
all files (which doesn't make sense because they are part of the
repository).
Adds additional messages to the error log when fetching contents. The
data is helpful in finding issues with receiving contents from servers.
References: #879, #882, #884
* [Exceptions] Don't return header for bridge exceptions
* [Exceptions] Add link to list in exception message
This is an alternative when the button is not rendered
for some reason.
* [index] Don't return bridge exception for formats
* [index] Return feed item for bridge exceptions
* [BridgeAbstract] Rename 'getCacheTime' to 'getModifiedTime'
* [BridgeAbstract] Move caching to index.php to separate concerns
index.php needs more control over caching behavior in order to cache
exceptions. This cannot be done in a bridge, as the bridge might be
broken, thus preventing caching from working.
This also (and more importantly) separates concerns. The bridge should
not even care if caching is involved or not. Its purpose is to collect
and provide data.
Response times should be faster, as more complex bridge functions like
'setDatas' (evaluates all input parameters to predict the current
context) and 'collectData' (collects data from sites) can be skipped
entirely.
Notice: In its current form, index.php takes care of caching. This
could, however, be moved into a separate class (i.e. CacheAbstract)
in order to make implementation details cache specific.
* [index] Add '_error_time' parameter to $item['uri']
This ensures that error messages are recognized by feed readers as
new errors after 24 hours. During that time the same item is returned
no matter how often the cache is cleared.
References https://github.com/RSS-Bridge/rss-bridge/issues/814#issuecomment-420876162
* [index] Include '_error_time' in the title for errors
This prevents feed readers from "updating" feeds based on the title
* [index] Handle "HTTP_IF_MODIFIED_SINCE" client requests
Implementation is based on `BridgeAbstract::dieIfNotModified()`,
introduced in 422c125d8e and
simplified based on https://stackoverflow.com/a/10847262
Basically, before returning cached data we check if the client send
the "HTTP_IF_MODIFIED_SINCE" header. If the modification time is
more recent or equal to the cache time, we reply with "HTTP/1.1 304
Not Modified" (same as before). Otherwise send the cached data.
* [index] Don't encode exception message with `htmlspecialchars`
* [Exceptions] Include error message in exception
* [index] Show different error message for error code 0
Adds a new class 'ParameterValidator' to replace the functions from
'validator.php', separating private functions from 'validateData' to
class private functions in the process.
Instead of echoing error messages, adds messages to a private variable,
accessible via 'getInvalidParameters'.
BridgeAbstract now adds invalid parameter names to the error message.
