morss/README.md

396 lines
13 KiB
Markdown
Raw Normal View History

# Morss - Get full-text RSS feeds
2013-02-25 15:40:51 +00:00
2020-09-03 11:17:58 +00:00
_GNU AGPLv3 code_
_Provided logo is CC BY-NC-SA 4.0_
2018-11-11 15:13:32 +00:00
2020-03-19 22:06:25 +00:00
Upstream source code: https://git.pictuga.com/pictuga/morss
Github mirror (for Issues & Pull requests): https://github.com/pictuga/morss
2020-03-19 22:04:21 +00:00
Homepage: https://morss.it/
This tool's goal is to get full-text RSS feeds out of striped RSS feeds,
commonly available on internet. Indeed most newspapers only make a small
description available to users in their rss feeds, which makes the RSS feed
rather useless. So this tool intends to fix that problem.
2013-02-25 20:49:38 +00:00
This tool opens the links from the rss feed, then downloads the full article
from the newspaper website and puts it back in the rss feed.
Morss also provides additional features, such as: .csv and json export, extended
control over output. A strength of morss is its ability to deal with broken
feeds, and to replace tracking links with direct links to the actual content.
2020-10-01 20:25:53 +00:00
Morss can also generate feeds from html and json files (see `feeds.py`), which
for instance makes it possible to get feeds for Facebook or Twitter, using
hand-written rules (ie. there's no automatic detection of links to build feeds).
Please mind that feeds based on html files may stop working unexpectedly, due to
html structure changes on the target website.
2020-04-05 20:20:33 +00:00
Additionally morss can detect rss feeds in html pages' `<meta>`.
2018-09-30 19:56:03 +00:00
You can use this program online for free at **[morss.it](https://morss.it/)**.
2018-09-30 19:56:30 +00:00
Some features of morss:
2020-10-01 20:25:53 +00:00
2018-09-30 19:56:30 +00:00
- Read RSS/Atom feeds
- Create RSS feeds from json/html pages
- Export feeds as RSS/JSON/CSV/HTML
- Fetch full-text content of feed items
- Follow 301/meta redirects
- Recover xml feeds with corrupt encoding
- Supports gzip-compressed http content
- HTTP caching with 3 different backends (in-memory/sqlite/mysql)
- Works as server/cli tool
- Deobfuscate various tracking links
2020-10-01 20:25:53 +00:00
## Install
### Python package
```shell
pip install git+https://git.pictuga.com/pictuga/morss.git
```
The dependency `lxml` is fairly long to install (especially on Raspberry Pi, as
C code needs to be compiled). If possible on your distribution, try installing
it with the system package manager.
2020-10-01 20:25:53 +00:00
Dependencies:
2015-04-07 11:01:41 +00:00
- [python](http://www.python.org/) >= 2.6 (python 3 is supported)
- [lxml](http://lxml.de/) for xml parsing
- [bs4](https://pypi.org/project/bs4/) for badly-formatted html pages
2013-10-12 21:43:09 +00:00
- [dateutil](http://labix.org/python-dateutil) to parse feed dates
2017-03-08 21:45:13 +00:00
- [chardet](https://pypi.python.org/pypi/chardet)
- [six](https://pypi.python.org/pypi/six), a dependency of chardet
- pymysql
2020-10-01 20:25:53 +00:00
You may also need:
- Apache, with python-cgi support, to run on a server
- a fast internet connection
### Docker
Build & run
2020-03-20 10:33:52 +00:00
```shell
docker build --tag morss https://git.pictuga.com/pictuga/morss.git --no-cache --pull
2020-10-01 20:25:53 +00:00
docker run -p 8080:8080 morss
2020-03-20 10:33:52 +00:00
```
2020-10-01 20:25:53 +00:00
With docker-compose:
2020-10-01 20:25:53 +00:00
```yml
services:
app:
build: https://git.pictuga.com/pictuga/morss.git
image: morss
ports:
- '8080:8080'
```
2020-10-01 20:25:53 +00:00
Then execute
2020-10-01 20:25:53 +00:00
```shell
docker-compose build --no-cache --pull
2020-10-01 20:25:53 +00:00
docker-compose up
```
2020-10-01 20:25:53 +00:00
## Run
morss will auto-detect what "mode" to use.
### Running on/as a server
Set up the server as indicated below, then visit:
```
2020-10-01 20:25:53 +00:00
http://PATH/TO/MORSS/[main.py/][:argwithoutvalue[:argwithvalue=value[...]]]/FEEDURL
```
2020-10-01 20:25:53 +00:00
For example: `http://morss.example/:clip/https://twitter.com/pictuga`
2020-10-01 20:25:53 +00:00
*(Brackets indicate optional text)*
2020-10-01 20:25:53 +00:00
The `main.py` part is only needed if your server doesn't support the Apache
redirect rule set in the provided `.htaccess`.
2020-10-01 20:25:53 +00:00
Works like a charm with [Tiny Tiny
RSS](http://tt-rss.org/redmine/projects/tt-rss/wiki), and most probably other
clients.
2020-10-01 20:25:53 +00:00
#### Via Docker
2020-10-01 20:25:53 +00:00
See above (in Install)
2020-10-01 20:25:53 +00:00
#### Using Gunicorn
```shell
gunicorn --preload morss
```
2020-10-01 20:25:53 +00:00
#### Using uWSGI
2020-10-01 20:25:53 +00:00
Running this command should do:
2013-04-28 09:37:11 +00:00
2020-10-01 20:25:53 +00:00
```shell
uwsgi --http :8080 --plugin python --wsgi-file main.py
```
#### Using morss' internal HTTP server
Morss can run its own, **very basic**, HTTP server, meant for debugging mostly.
The latter should start when you run morss without any argument, on port 8080.
I'd highly recommend you to use gunicorn or something similar for better
performance.
```shell
morss
```
You can change the port using environment variables like this `PORT=9000 morss`.
2013-04-28 09:37:11 +00:00
#### Via mod_cgi/FastCGI with Apache/nginx
For this, you'll want to change a bit the architecture of the files, for example
into something like this.
```
/
├── cgi
│   │
│   ├── main.py
│   ├── morss
│   │   ├── __init__.py
│   │   ├── __main__.py
│   │   ├── morss.py
│   │   └── ...
│   │
│   ├── dateutil
│   └── ...
├── .htaccess
├── index.html
└── ...
```
For this, you need to make sure your host allows python script execution. This
method uses HTTP calls to fetch the RSS feeds, which will be handled through
`mod_cgi` for example on Apache severs.
Please pay attention to `main.py` permissions for it to be executable. Also
ensure that the provided `/www/.htaccess` works well with your server.
### As a CLI application
2020-03-20 10:41:43 +00:00
Run:
2020-10-01 20:25:53 +00:00
2020-03-20 10:41:43 +00:00
```
morss [--argwithoutvalue] [--argwithvalue=value] [...] FEEDURL
2020-03-20 10:41:43 +00:00
```
2020-10-01 20:25:53 +00:00
2020-08-24 17:39:27 +00:00
For example: `morss --clip http://feeds.bbci.co.uk/news/rss.xml`
2020-03-20 10:41:43 +00:00
*(Brackets indicate optional text)*
### As a newsreader hook
To use it, the newsreader [Liferea](http://lzone.de/liferea/) is required
(unless other newsreaders provide the same kind of feature), since custom
scripts can be run on top of the RSS feed, using its
[output](http://lzone.de/liferea/scraping.htm) as an RSS feed.
2020-10-01 20:25:53 +00:00
To use this script, you have to enable "(Unix) command" in liferea feed
settings, and use the command:
2020-03-20 10:41:43 +00:00
```
morss [argwithoutvalue] [argwithvalue=value] [...] FEEDURL
2020-03-20 10:41:43 +00:00
```
2020-10-01 20:25:53 +00:00
For example: `morss http://feeds.bbci.co.uk/news/rss.xml`
2020-03-20 10:41:43 +00:00
*(Brackets indicate optional text)*
### As a python library
2014-05-24 17:03:05 +00:00
Quickly get a full-text feed:
2020-10-01 20:25:53 +00:00
2014-05-24 17:03:05 +00:00
```python
>>> import morss
2014-05-24 17:14:12 +00:00
>>> xml_string = morss.process('http://feeds.bbci.co.uk/news/rss.xml')
2014-05-24 17:03:05 +00:00
>>> xml_string[:50]
"<?xml version='1.0' encoding='UTF-8'?>\n<?xml-style"
```
2014-05-24 17:14:12 +00:00
Using cache and passing arguments:
2020-10-01 20:25:53 +00:00
2014-05-24 17:03:05 +00:00
```python
>>> import morss
>>> url = 'http://feeds.bbci.co.uk/news/rss.xml'
>>> cache = '/tmp/morss-cache.db' # sqlite cache location
>>> options = {'csv':True}
2014-05-24 17:03:05 +00:00
>>> xml_string = morss.process(url, cache, options)
>>> xml_string[:50]
'{"title": "BBC News - Home", "desc": "The latest s'
```
`morss.process` is actually a wrapper around simpler function. It's still
possible to call the simpler functions, to have more control on what's happening
under the hood.
2014-05-24 17:03:05 +00:00
Doing it step-by-step:
2020-10-01 20:25:53 +00:00
```python
import morss, morss.crawler
2014-05-21 13:05:35 +00:00
url = 'http://newspaper.example/feed.xml'
options = morss.Options(csv=True) # arguments
morss.crawler.sqlite_default = '/tmp/morss-cache.db' # sqlite cache location
url, rss = morss.FeedFetch(url, options) # this only grabs the RSS feed
2017-03-09 01:17:40 +00:00
rss = morss.FeedGather(rss, url, options) # this fills the feed and cleans it up
2020-04-27 16:12:14 +00:00
output = morss.FeedFormat(rss, options, 'unicode') # formats final feed
```
2020-10-01 20:25:53 +00:00
## Arguments and settings
2020-10-01 20:25:53 +00:00
### Arguments
2020-08-23 16:43:18 +00:00
2020-10-01 20:25:53 +00:00
morss accepts some arguments, to lightly alter the output of morss. Arguments
2020-10-03 18:56:40 +00:00
may need to have a value (usually a string or a number). How to pass those
arguments to morss is explained in Run above.
2020-10-01 20:25:53 +00:00
The list of arguments can be obtained by running `morss --help`
```
usage: morss [-h] [--format {rss,json,html,csv}] [--search STRING] [--clip]
[--indent] [--cache] [--force] [--proxy] [--newest] [--firstlink]
[--resolve] [--items XPATH] [--item_link XPATH]
[--item_title XPATH] [--item_content XPATH] [--item_time XPATH]
[--nolink] [--noref] [--silent]
url
Get full-text RSS feeds
positional arguments:
url feed url
2020-10-01 20:25:53 +00:00
optional arguments:
-h, --help show this help message and exit
output:
--format {rss,json,html,csv}
output format
--search STRING does a basic case-sensitive search in the feed
--clip stick the full article content under the original feed
content (useful for twitter)
--indent returns indented XML or JSON, takes more place, but
human-readable
action:
--cache only take articles from the cache (ie. don't grab new
articles' content), so as to save time
--force force refetch the rss feed and articles
--proxy doesn't fill the articles
--newest return the feed items in chronological order (morss
ohterwise shows the items by appearing order)
--firstlink pull the first article mentioned in the description
instead of the default link
--resolve replace tracking links with direct links to articles
(not compatible with --proxy)
custom feeds:
--items XPATH (mandatory to activate the custom feeds function)
xpath rule to match all the RSS entries
--item_link XPATH xpath rule relative to items to point to the entry's
link
--item_title XPATH entry's title
--item_content XPATH entry's content
--item_time XPATH entry's date & time (accepts a wide range of time
formats)
misc:
--nolink drop links, but keeps links' inner text
--noref drop items' link
--silent don't output the final RSS (useless on its own, but
can be nice when debugging)
GNU AGPLv3 code
```
Further HTTP-only options:
- `callback=NAME`: for JSONP calls
- `cors`: allow Cross-origin resource sharing (allows XHR calls from other
servers)
- `txt`: changes the http content-type to txt (for faster "`view-source:`")
### Environment variables
To pass environment variables:
- Docker-cli: `docker run -p 8080:8080 morss --env KEY=value`
- docker-compose: add an `environment:` section in the .yml file
- Gunicorn/uWSGI/CLI: prepend `KEY=value` before the command
- Apache: via the `SetEnv` instruction (see sample `.htaccess` provided)
Generic:
2020-10-03 18:56:40 +00:00
- `DEBUG=1`: to have some feedback from the script execution. Useful for
debugging.
2020-10-03 17:57:08 +00:00
- `IGNORE_SSL=1`: to ignore SSL certs when fetch feeds and articles
- `DELAY` (seconds) sets the browser cache delay, only for HTTP clients
- `TIMEOUT` (seconds) sets the HTTP timeout when fetching rss feeds and articles
When parsing long feeds, with a lot of items (100+), morss might take a lot of
time to parse it, or might even run into a memory overflow on some shared
hosting plans (limits around 10Mb), in which case you might want to adjust the
2020-08-23 14:09:58 +00:00
below settings via environment variables.
2013-05-23 19:48:45 +00:00
- `MAX_TIME` (seconds) sets the maximum amount of time spent *fetching*
articles, more time might be spent taking older articles from cache. `-1` for
unlimited.
2020-10-01 20:25:53 +00:00
- `MAX_ITEM` sets the maximum number of articles to fetch. `-1` for unlimited.
More articles will be taken from cache following the nexts settings.
- `LIM_TIME` (seconds) sets the maximum amount of time spent working on the feed
(whether or not it's already cached). Articles beyond that limit will be dropped
from the feed. `-1` for unlimited.
2020-10-01 20:25:53 +00:00
- `LIM_ITEM` sets the maximum number of article checked, limiting both the
number of articles fetched and taken from cache. Articles beyond that limit will
be dropped from the feed, even if they're cached. `-1` for unlimited.
2020-10-01 20:25:53 +00:00
morss uses caching to make loading faster. There are 3 possible cache backends:
2020-10-01 20:25:53 +00:00
- `(nothing/default)`: a simple python in-memory dict-like object.
- `CACHE=sqlite`: sqlite3 cache. Default file location is in-memory (i.e. it
will be cleared every time the program is run). Path can be defined with
`SQLITE_PATH`.
- `CACHE=mysql`: MySQL cache. Connection can be defined with the following
environment variables: `MYSQL_USER`, `MYSQL_PWD`, `MYSQL_DB`, `MYSQL_HOST`
To limit the size of the cache:
- `CACHE_SIZE` sets the target number of items in the cache (further items will
be deleted but the cache might be temporarily bigger than that). Defaults to 1k
2020-10-01 20:25:53 +00:00
entries.
- `CACHE_LIFESPAN` (seconds) sets how often the cache must be trimmed (i.e. cut
down to the number of items set in `CACHE_SIZE`). Defaults to 1min.
### Content matching
The content of articles is grabbed with our own readability fork. This means
that most of the time the right content is matched. However sometimes it fails,
therefore some tweaking is required. Most of the time, what has to be done is to
2020-10-01 20:25:53 +00:00
add some "rules" in the main script file in `readabilite.py` (not in morss).
2013-03-29 19:05:53 +00:00
Most of the time when hardly nothing is matched, it means that the main content
of the article is made of images, videos, pictures, etc., which readability
doesn't detect. Also, readability has some trouble to match content of very
small articles.
2013-04-04 16:31:26 +00:00
morss will also try to figure out whether the full content is already in place
(for those websites which understood the whole point of RSS feeds). However this
detection is very simple, and only works if the actual content is put in the
"content" section in the feed and not in the "summary" section.