pferd/CONFIG.md

366 lines
15 KiB
Markdown
Raw Normal View History

2021-04-29 18:55:08 +02:00
# Config file format
A config file consists of sections. A section begins with a `[section]` header,
2021-05-25 14:12:19 +02:00
which is followed by a list of `key = value` pairs. Comments must be on their
own line and start with `#`. Multiline values must be indented beyond their key.
Boolean values can be `yes` or `no`. For more details and some examples on the
format, see the [configparser documentation][1] ([basic interpolation][2] is
enabled).
2021-04-29 18:55:08 +02:00
[1]: <https://docs.python.org/3/library/configparser.html#supported-ini-file-structure> "Supported INI File Structure"
[2]: <https://docs.python.org/3/library/configparser.html#configparser.BasicInterpolation> "BasicInterpolation"
## The `DEFAULT` section
This section contains global configuration values. It can also be used to set
default values for the other sections.
- `working_dir`: The directory PFERD operates in. Set to an absolute path to
2021-05-25 14:12:19 +02:00
make PFERD operate the same regardless of where it is executed from. All other
2021-04-29 18:55:08 +02:00
paths in the config file are interpreted relative to this path. If this path
is relative, it is interpreted relative to the script's working dir. `~` is
expanded to the current user's home directory. (Default: `.`)
- `explain`: Whether PFERD should log and explain its actions and decisions in
detail. (Default: `no`)
2021-05-25 14:12:19 +02:00
- `status`: Whether PFERD should print status updates (like `Crawled ...`,
`Added ...`) while running a crawler. (Default: `yes`)
2021-05-23 22:51:42 +02:00
- `report`: Whether PFERD should print a report of added, changed and deleted
local files for all crawlers before exiting. (Default: `yes`)
2021-05-25 14:12:19 +02:00
- `share_cookies`: Whether crawlers should share cookies where applicable. For
example, some crawlers share cookies if they crawl the same website using the
same account. (Default: `yes`)
2021-04-29 18:55:08 +02:00
## The `crawl:*` sections
Sections whose names start with `crawl:` are used to configure crawlers. The
rest of the section name specifies the name of the crawler.
A crawler synchronizes a remote resource to a local directory. There are
different types of crawlers for different kinds of resources, e. g. ILIAS
courses or lecture websites.
Each crawl section represents an instance of a specific type of crawler. The
`type` option is used to specify the crawler type. The crawler's name is usually
2021-05-25 14:12:19 +02:00
used as the output directory. New crawlers can be created simply by adding a new
crawl section to the config file.
2021-04-29 18:55:08 +02:00
Depending on a crawler's type, it may have different options. For more details,
2021-05-25 14:12:19 +02:00
see the type's [documentation](#crawler-types) below. The following options are
common to all crawlers:
2021-04-29 18:55:08 +02:00
2021-05-25 14:12:19 +02:00
- `type`: The available types are specified in [this section](#crawler-types).
2021-04-29 18:55:08 +02:00
- `output_dir`: The directory the crawler synchronizes files to. A crawler will
never place any files outside of this directory. (Default: the crawler's name)
2021-05-25 14:12:19 +02:00
- `redownload`: When to download a file that is already present locally.
2021-04-30 15:32:56 +02:00
(Default: `never-smart`)
- `never`: If a file is present locally, it is not downloaded again.
- `never-smart`: Like `never`, but PFERD tries to detect if an already
downloaded files has changed via some (unreliable) heuristics.
- `always`: All files are always downloaded, regardless of whether they are
already present locally.
- `always-smart`: Like `always`, but PFERD tries to avoid unnecessary
downloads via some (unreliable) heuristics.
2021-05-05 00:55:55 +02:00
- `on_conflict`: What to do when the local and remote versions of a file or
2021-05-25 14:12:19 +02:00
directory differ, including when a file is replaced by a directory or a
directory by a file. (Default: `prompt`)
2021-05-05 00:55:55 +02:00
- `prompt`: Always ask the user before overwriting or deleting local files
and directories.
- `local-first`: Always keep the local file or directory. Equivalent to
using `prompt` and always choosing "no". Implies that `redownload` is set
to `never`.
- `remote-first`: Always keep the remote file or directory. Equivalent to
using `prompt` and always choosing "yes".
- `no-delete`: Never delete local files, but overwrite local files if the
remote file is different.
2021-04-29 18:55:08 +02:00
- `transform`: Rules for renaming and excluding certain files and directories.
For more details, see [this section](#transformation-rules). (Default: empty)
2021-05-25 14:12:19 +02:00
- `tasks`: The maximum number of concurrent tasks (such as crawling or
downloading). (Default: `1`)
- `downloads`: How many of those tasks can be download tasks at the same time.
Must not be greater than `tasks`. (Default: Same as `tasks`)
- `task_delay`: Time (in seconds) that the crawler should wait between
subsequent tasks. Can be used as a sort of rate limit to avoid unnecessary
2021-05-25 14:12:19 +02:00
load for the crawl target. (Default: `0.0`)
- `windows_paths`: Whether PFERD should find alternative names for paths that
are invalid on Windows. (Default: `yes` on Windows, `no` otherwise)
2021-04-29 18:55:08 +02:00
2021-05-09 01:33:47 +02:00
Some crawlers may also require credentials for authentication. To configure how
the crawler obtains its credentials, the `auth` option is used. It is set to the
full name of an auth section (including the `auth:` prefix).
Here is a simple example:
2021-05-25 17:16:57 +02:00
```ini
2021-05-09 01:33:47 +02:00
[auth:example]
type = simple
username = foo
password = bar
[crawl:something]
type = some-complex-crawler
auth = auth:example
2021-05-25 14:12:19 +02:00
on_conflict = no-delete
tasks = 3
2021-05-09 01:33:47 +02:00
```
2021-04-29 18:55:08 +02:00
## The `auth:*` sections
Sections whose names start with `auth:` are used to configure authenticators. An
authenticator provides a username and a password to one or more crawlers.
2021-04-29 18:55:08 +02:00
Authenticators work similar to crawlers: A section represents an authenticator
2021-05-25 14:12:19 +02:00
instance whose name is the rest of the section name. The type is specified by
2021-04-29 18:55:08 +02:00
the `type` option.
Depending on an authenticator's type, it may have different options. For more
2021-05-25 14:12:19 +02:00
details, see the type's [documentation](#authenticator-types) below. The only
option common to all authenticators is `type`:
2021-04-29 18:55:08 +02:00
- `type`: The types are specified in [this section](#authenticator-types).
## Crawler types
2021-05-09 01:33:47 +02:00
### The `local` crawler
This crawler crawls a local directory. It is really simple and mostly useful for
testing different setups. The various delay options are meant to make the
crawler simulate a slower, network-based crawler.
2021-05-09 01:33:47 +02:00
2021-05-15 17:12:25 +02:00
- `target`: Path to the local directory to crawl. (Required)
2021-05-25 14:12:19 +02:00
- `crawl_delay`: Artificial delay (in seconds) to simulate for crawl requests.
(Default: `0.0`)
- `download_delay`: Artificial delay (in seconds) to simulate for download
requests. (Default: `0.0`)
- `download_speed`: Download speed (in bytes per second) to simulate. (Optional)
2021-04-29 18:55:08 +02:00
2021-05-25 14:12:19 +02:00
### The `kit-ilias-web` crawler
This crawler crawls the KIT ILIAS instance.
ILIAS is not great at handling too many concurrent requests. To avoid
unnecessary load, please limit `tasks` to `1`.
There is a spike in ILIAS usage at the beginning of lectures, so please don't
run PFERD during those times.
If you're automatically running PFERD periodically (e. g. via cron or a systemd
timer), please randomize the start time or at least don't use the full hour. For
systemd timers, this can be accomplished using the `RandomizedDelaySec` option.
Also, please schedule the script to run in periods of low activity. Running the
script once per day should be fine.
- `target`: The ILIAS element to crawl. (Required)
- `desktop`: Crawl your personal desktop
- `<course id>`: Crawl the course with the given id
- `<url>`: Crawl a given element by URL (preferably the permanent URL linked
at the bottom of its ILIAS page)
- `auth`: Name of auth section to use for login. (Required)
- `tfa_auth`: Name of auth section to use for two-factor authentication. Only
uses the auth section's password. (Default: Anonymous `tfa` authenticator)
- `links`: How to represent external links. (Default: `fancy`)
- `ignore`: Don't download links.
- `plaintext`: A text file containing only the URL.
- `fancy`: A HTML file looking like the ILIAS link element.
- `internet-shortcut`: An internet shortcut file (`.url` file).
- `link_redirect_delay`: Time (in seconds) until `fancy` link files will
redirect to the actual URL. Set to a negative value to disable the automatic
redirect. (Default: `-1`)
- `videos`: Whether to download videos. (Default: `no`)
- `http_timeout`: The timeout (in seconds) for all HTTP requests. (Default:
`20.0`)
2021-05-23 23:40:28 +02:00
2021-04-29 18:55:08 +02:00
## Authenticator types
2021-05-13 19:55:04 +02:00
### The `simple` authenticator
With this authenticator, the username and password can be set directly in the
config file. If the username or password are not specified, the user is prompted
via the terminal.
- `username`: The username. (Optional)
- `password`: The password. (Optional)
2021-04-29 18:55:08 +02:00
2021-05-25 14:12:19 +02:00
### The `keyring` authenticator
This authenticator uses the system keyring to store passwords. The username can
be set directly in the config file. If the username is not specified, the user
is prompted via the terminal. If the keyring contains no entry or the entry is
incorrect, the user is prompted for a password via the terminal and the password
is stored in the keyring.
- `username`: The username. (Optional)
- `keyring_name`: The service name PFERD uses for storing credentials. (Default:
`PFERD`)
2021-05-15 18:27:16 +02:00
### The `tfa` authenticator
This authenticator prompts the user on the console for a two-factor
authentication token. The token is provided as password and it is not cached.
This authenticator does not support usernames.
2021-04-29 18:55:08 +02:00
## Transformation rules
Transformation rules are rules for renaming and excluding files and directories.
They are specified line-by-line in a crawler's `transform` option. When a
crawler needs to apply a rule to a path, it goes through this list top-to-bottom
and choose the first matching rule.
2021-05-26 11:47:51 +02:00
To see this process in action, you can use the `--debug-transforms` or flag or
the `--explain` flag.
2021-04-29 18:55:08 +02:00
Each line has the format `SOURCE ARROW TARGET` where `TARGET` is optional.
`SOURCE` is either a normal path without spaces (e. g. `foo/bar`), or a string
literal delimited by `"` or `'` (e. g. `"foo\" bar/baz"`). Python's string
escape syntax is supported. Trailing slashes are ignored. `TARGET` can be
formatted like `SOURCE`, but it can also be a single exclamation mark without
quotes (`!`). `ARROW` is one of `-->`, `-name->`, `-exact->`, `-re->` and
2021-05-15 15:06:45 +02:00
`-name-re->`
2021-04-29 18:55:08 +02:00
If a rule's target is `!`, this means that when the rule matches on a path, the
corresponding file or directory is ignored. If a rule's target is missing, the
path is matched but not modified.
### The `-->` arrow
The `-->` arrow is a basic renaming operation. If a path begins with `SOURCE`,
that part of the path is replaced with `TARGET`. This means that the rule
`foo/bar --> baz` would convert `foo/bar` into `baz`, but also `foo/bar/xyz`
into `baz/xyz`. The rule `foo --> !` would ignore a directory named `foo` as
well as all its contents.
2021-05-15 15:06:45 +02:00
### The `-name->` arrow
The `-name->` arrow lets you rename files and directories by their name,
regardless of where they appear in the file tree. Because of this, its `SOURCE`
must not contain multiple path segments, only a single name. This restriction
does not apply to its `TARGET`. The `-name->` arrow is not applied recursively
to its own output to prevent infinite loops.
For example, the rule `foo -name-> bar/baz` would convert `a/foo` into
`a/bar/baz` and `a/foo/b/c/foo` into `a/bar/baz/b/c/bar/baz`. The rule `foo
-name-> !` would ignore all directories and files named `foo`.
2021-05-15 15:06:45 +02:00
2021-04-29 18:55:08 +02:00
### The `-exact->` arrow
The `-exact->` arrow requires the path to match `SOURCE` exactly. This means
that the rule `foo/bar -exact-> baz` would still convert `foo/bar` into `baz`,
but `foo/bar/xyz` would be unaffected. Also, `foo -exact-> !` would only ignore
`foo`, but not its contents (if it has any). The examples below show why this is
useful.
### The `-re->` arrow
The `-re->` arrow uses regular expressions. `SOURCE` is a regular expression
that must match the entire path. If this is the case, then the capturing groups
are available in `TARGET` for formatting.
`TARGET` uses Python's [format string syntax][3]. The *n*-th capturing group can
be referred to as `{g<n>}` (e. g. `{g3}`). `{g0}` refers to the original path.
If capturing group *n*'s contents are a valid integer, the integer value is
available as `{i<n>}` (e. g. `{i3}`). If capturing group *n*'s contents are a
2021-05-27 13:56:01 +02:00
valid float, the float value is available as `{f<n>}` (e. g. `{f3}`). If a
capturing group is not present (e. g. when matching the string `cd` with the
regex `(ab)?cd`), the corresponding variables are not defined.
Python's format string syntax has rich options for formatting its arguments. For
example, to left-pad the capturing group 3 with the digit `0` to width 5, you
can use `{i3:05}`.
PFERD even allows you to write entire expressions inside the curly braces, for
example `{g2.lower()}` or `{g3.replace(' ', '_')}`.
[3]: <https://docs.python.org/3/library/string.html#format-string-syntax> "Format String Syntax"
2021-05-15 15:06:45 +02:00
### The `-name-re->` arrow
The `-name-re>` arrow is like a combination of the `-name->` and `-re->` arrows.
Instead of the `SOURCE` being the name of a directory or file, it's a regex that
is matched against the names of directories and files. `TARGET` works like the
`-re->` arrow's target.
For example, the arrow `(.*)\.jpeg -name-re-> {g1}.jpg` will rename all `.jpeg`
extensions into `.jpg`. The arrow `\..+ -name-re-> !` will ignore all files and
directories starting with `.`.
2021-05-15 15:06:45 +02:00
2021-04-29 18:55:08 +02:00
### Example: Tutorials
2021-05-05 00:55:55 +02:00
You have an ILIAS course with lots of tutorials, but are only interested in a
2021-05-15 15:06:45 +02:00
single one.
2021-04-29 18:55:08 +02:00
```
tutorials/
|- tut_01/
|- tut_02/
|- tut_03/
...
```
You can use a mix of normal and exact arrows to get rid of the other ones and
move the `tutorials/tut_02/` folder to `my_tut/`:
```
tutorials/tut_02 --> my_tut
tutorials -exact->
tutorials --> !
```
The second rule is required for many crawlers since they use the rules to decide
which directories to crawl. If it was missing when the crawler looks at
`tutorials/`, the third rule would match. This means the crawler would not crawl
the `tutorials/` directory and thus not discover that `tutorials/tut02/`
existed.
Since the second rule is only relevant for crawling, the `TARGET` is left out.
### Example: Lecture slides
You have a course with slides like `Lecture 3: Linear functions.PDF` and you
would like to rename them to `03_linear_functions.pdf`.
```
Lectures/
|- Lecture 1: Introduction.PDF
|- Lecture 2: Vectors and matrices.PDF
|- Lecture 3: Linear functions.PDF
...
```
2021-05-05 00:55:55 +02:00
To do this, you can use the most powerful of arrows: The regex arrow.
```
"Lectures/Lecture (\\d+): (.*)\\.PDF" -re-> "Lectures/{i1:02}_{g2.lower().replace(' ', '_')}.pdf"
```
Note the escaped backslashes on the `SOURCE` side.
2021-05-15 15:06:45 +02:00
### Example: Crawl a python project
You are crawling a python project and want to ignore all hidden files (files
whose name starts with a `.`), all `__pycache__` directories and all markdown
files (for some weird reason).
```
.gitignore
.mypy_cache/
.venv/
CONFIG.md
PFERD/
|- __init__.py
|- __main__.py
|- __pycache__/
|- authenticator.py
|- config.py
...
README.md
...
```
For this task, the name arrows can be used. They are variants of the normal
arrows that only look at the file name instead of the entire path.
```
\..* -name-re-> !
__pycache__ -name-> !
.*\.md -name-re-> !
2021-05-15 15:06:45 +02:00
```