Config Options¶

Note

Default Value is specified by parenthesis in the first lines.

Value Function is specified by bracket in the first lines.

tosixinch.ini¶

Note

Options with star * are common options with site.ini. You can use them to override application-wide configuration.

General Section¶

downloader *¶

(urllib)

Specify default downloader. Either urllib or headless.

extractor *¶

(lxml)

Specify default extractor. Currently the only option is lxml.

converter¶

(prince)

Specify default converter. prince or weasyprint.

user_agent *¶

(some arbitrary browser user agent. Run 'tosixinch -a' to actually see.)

Specify user agent for downloader (only for urllib).

browser_engine *¶

(selenium-firefox)

Specify browser engine when headless option is True, selenium-chrome or selenium-firefox.

selenium_chrome_path *¶: (None)

Specify the path of chromedriver for selenium (passed to executable_path argument). Normally unnecessary.

selenium_firefox_path *¶: (None)

Specify the path of geckodriver for selenium (passed to executable_path argument). Normally unnecessary.

encoding *¶

(utf-8, cp1252, latin_1)

[COMMA]

Specify preferred encoding or encodings. First successful one is used. Encoding names are as specified in codecs library, or ‘html5prescan’, or ‘ftfy’ if they are installed.

If the name is html5prescan, html5prescan tries to get a valid encoding declaration from html. (The library strictly follows html5 spec and usually it is not necessary nor useful. It is intended for occasional debug purpose.)

After successful encoding by one of the encodings, if the list includes ftfy, ftfy.fixes.fix_encoding method is called with the decoded text. It may be able to fix some ‘mojibake’. (So it is always called last, the place in the list is irrelevant).

Note

The included bash completion only completes canonical codec names (with underline changed to dash). But you can put any other name or alias names, as long as they are legal in your Python environment.

encoding_errors *¶

(strict)

Specify codec Error Handler.

If you can’t run extract because of decoding errors, one solution is to change this option to ‘replace’ or ‘backslashreplace’.

parts_download *¶

(True)

[BOOL]

Web pages may have some component content. Most important ones are images, and currently the program only concerns images (in html tag <img src=...>). The value specifies whether it downloads these components when extract.

Note downloading may occur anyway by pdf converters.

If this option is True, download links are rewritten to point to local dfiles. So downloading doesn’t happen when convert.

In general, pre-downloading is useful for multiple trials and layout checking.

If force_download is False (default), the program skips downloading if the file already exists.

force_download *¶

(False)

[BOOL]

By default, The program does not download if the destination file exists.

If this options is True:

In case of -1, it (re-) downloads URL even if dfile exists.

In case of -2, it (re-) downloads component files (images etc.) even if they exist.

But in one invocation, this re-downloading is always once for one URL. (The program doesn’t download the same icon files again and again).

guess¶

(//div[@itemprop="articleBody"]
//div[@role="main"]
//div[@id="main"]
//div[@id="content"]
//div[@class=="body"]
//article)

[LINE]

If html rsrc doesn’t match any match option in site.ini, select is done according to this value.

The procedure:

The XPaths in this value are searched in order, line by line.
If match is found and match is a single element (not multiple occurrences), the program select s the element.

defaultprocess *¶

(add_h1, youtube_video_to_thumbnail, convert_permalink_sign)

[LINE]

Before site specific process functions, the program applies default process functions to all html rsrc, according to this value.

The syntax is the same as process option, in site.ini.

About default functions:

add_h1: If there is no <h1>, make <h1> tag from <title> tag text. It is to make better pdf bookmarks (TOC).

youtube_video_to_thumbnail: Change embedded youtube video object to thumbnail image.

convert_permalink_sign: Remove permalink sign (’¶’), for a few class (‘headerlink’ etc.). Python documents tend to use them, and On pdf, they are always visible, rather noisy.

When the default functions is undesirable in some site, please override this option in user site.ini.

full_image *¶

(200)

[INT]

If width or height of image pixel size is equal or above this value, class attribute tsi-tall or tsi-wide is added to the image tag, tsi-tall if height/width ratio is greater than the ratio of the e-reader display, tsi-wide if the opposite.

By itself, it does nothing. However, In sample.css, it is used to make medium sized images expand almost full display size, with small images (icon, logo, etc.) intact. The layout gets a bit uglier, but I think it is necessary for small e-reader displays.

add_binary_extensions¶

(3dm 3ds 3g2 3gp 7z a aac adp ai aif aiff alz ape apk appimage ar arj asf au avi bak baml bh bin bk bmp btif bz2 bzip2 cab caf cgm class cmx cpio cr2 cur dat dcm deb dex djvu dll dmg dng doc docm docx dot dotm dra DS_Store dsk dts dtshd dvb dwg dxf ecelp4800 ecelp7470 ecelp9600 egg eol eot epub exe f4v fbs fh fla flac flatpak fli flv fpx fst fvt g3 gh gif graffle gz gzip h261 h263 h264 icns ico ief img ipa iso jar jpeg jpg jpgv jpm jxr key ktx lha lib lvp lz lzh lzma lzo m3u m4a m4v mar mdi mht mid midi mj2 mka mkv mmr mng mobi mov movie mp3 mp4 mp4a mpeg mpg mpga mxu nef npx numbers nupkg o odp ods odt oga ogg ogv otf ott pages pbm pcx pdb pdf pea pgm pic png pnm pot potm potx ppa ppam ppm pps ppsm ppsx ppt pptm pptx psd pya pyc pyo pyv qt rar ras raw resources rgb rip rlc rmf rmvb rpm rtf rz s3m s7z scpt sgi shar snap sil sketch slk smv snk so stl suo sub swf tar tbz tbz2 tga tgz thmx tif tiff tlz ttc ttf txz udf uvh uvi uvm uvp uvs uvu viv vob war wav wax wbmp wdp weba webm webp whl wim wm wma wmv wmx woff woff2 wrm wvx xbm xif xla xlam xls xlsb xlsm xlsx xlt xltm xltx xm xmind xpi xpm xwd xz z zip zipx)

[PLUS]

The program ignores rsrcs with binary like looking extensions, only when multiple rsrcs are provided.

This option value adds to or subtracts from the default add_binary_extensions list above.

The list is taken from Sindre Sorhus’ binary-extensions.

This is for user convenience. If you copy and paste many rsrcs, checking strange extensions is a bit of work. But I’m afraid sometimes it gets in the way.

(An example I found: some old unix software uses doc extension for text (like README.doc).

add_clean_tags *¶

(None)

[PLUS]

After select, exclude and process in extract, the program clean s the resultant html.

The tags in this option are stripped. The current default is none.

add_clean_attrs *¶

(color, width, height)

[PLUS]

After select, exclude and process in extract, the program clean s the resultant html.

The attributes in this option are stripped. The current default is color, width and height.

Most e-readers are black and white. Colors just make fonts harder to read.

Width and height conflict with user css rules.

elements_to_keep_attrs *¶

(self::math
self::svg
self::node()[starts-with(@class, "MathJax")])

[LINE]

After select, exclude and process in extract, the program clean s the resultant html.

The program skips cleaning attributes for the elements that matches one of the XPath in this option.

The default is math, svg and some MathJax related tags. They have inter-related width and height information, which we usually want to keep.

Note XPaths are checked from each element, not from the root document. So the selectors are like above (not like e.g. '//math').

styles_to_retain *¶

(content:;
display: none;
text-decoration-line: line-through;
text-decoration: line-through;)

[LINE]

Specify particular inline styles you want to retain (all other styles are removed).

If only the property is provided (no values after ':'), then all values are retained as is.

‘;’ before each line end is optional.

Note: This is for the simplest css manipulations. The program gives up for any indication of complexity (currently the existence of any characters of ‘,’, ‘/’, ‘(‘).

clean *¶

(both)

Specify how to clean html (both, head, body, none).

See Clean.

ftype¶

(None)

Specify file type when extract.

Valid values are:

'html', 'prose', 'nonprose', 'python'

It needs improvement.

textwidth¶

(65)

[INT]

Set physical line length for nonprose texts.

See nonprose.

textindent¶

(' --> ')

Set logical line continuation marker for nonprose texts.

See nonprose.

ConfigParser strips leading and ending whitespaces. So if you want actual whitespaces, quote them as the default does. Quotes are stripped by the program in turn.

trimdirs *¶

(3)

[INT]

Shorten PDF table of contents title, if it is a local text file.

PDF toc title for local text file is made from their full path. If this trimdirs option value is with no sign, remove that number from leading path segments. If it is with minus sign, remove leading path segments to make the segments to that number.

--trimdirs 0
aaa/bbb/ccc/ddd/eee/fff

--trimdirs 2  # remove two segments
ccc/ddd/eee/fff

--trimdirs -2  # reduce to two segments
eee/fff

# c.f. no bounding errors

--trimdirs 100
fff

--trimdirs -100
aaa/bbb/ccc/ddd/eee/fff

Note html files always use html title (actual, or placeholder notitle). Remote text (non-html) files use the URL with scheme (’https://’) stripped.

C.f. –check commandline option prints out this shortened names for local files. They include local html files, so it is not perfect, but it can be useful for checking and adjusting this trimdirs option.

raw¶: (False)

[BOOL]

If True, when convert, the program processes rsrcs. Normally (if it is False), it processes efile.

css *¶

(sample)

[COMMA]

CSS file names to be used in order. The names are referenced, in order, in efiles ('<link ... rel="stylesheet">').

you can only use the filenames (not full paths).

The filenames are searched in css directory, application css directory and current directory in order.

The program includes sample css sample.t.css, and as a special case, it can be abbreviated as sample (default).

pdfname¶: (None)

Specify output PDF file name. If not provided (default), the program makes up some name. see PDF_File.

—

Note

For hookcmds below, see Hookcmds.

precmd1¶: (None)

[LINE][CMDS]

Run arbitrary commands before download action.

postcmd1¶: (None)

[LINE][CMDS]

Run arbitrary commands after download action.

precmd2¶: (None)

[LINE][CMDS]

Run arbitrary commands before extract action.

postcmd2¶: (None)

[LINE][CMDS]

Run arbitrary commands after extract action.

precmd3¶: (None)

[LINE][CMDS]

Run arbitrary commands before convert action.

postcmd3¶: (None)

[LINE][CMDS]

Run arbitrary commands after convert action.

viewcmd¶: (None)

[LINE][CMDS]

Run arbitrary commands when specified in commandline options (-4 or --view).

pre_each_cmd1¶: (None)

[LINE][CMDS]

Run arbitrary commands before each download.

post_each_cmd1¶: (None)

[LINE][CMDS]

Run arbitrary commands after each download.

pre_each_cmd2¶

(None)

[LINE][CMDS]

Run arbitrary commands before each extract.

There are sample hook extractors. See _man and _pcode.

post_each_cmd2¶: (None)

[LINE][CMDS]

Run arbitrary commands after each extract.

browsercmd¶

(None)

[CMD]

When action is --browser, the default is just call Python stdlib webbrowser to open a browser. If it is not desirable, specify the open command here, e.g.:

firefox 'site.slash_efile'

You have to use the magic word site.slash_efile for the filename. It evaluates to the intended URL version of efile (percent encoding etc.).

Style Section¶

The options in style section are used for css template files.

Note that users can always choose (static) css files rather than css template files. In that case, the style options have no effect.

So, the options themselves have no meaning. In the following, the roles in the sample file (sample.t.css) are explained.

orientation¶

(portrait)

Specify page orientation, portrait or landscape.

portrait_size¶

(90mm 118mm)

Specify portrait page size (width and height). The program uses this value when orientation is portrait.

The display size of common 6-inch e-readers seems around 90mm x 120mm. Here the default thinly clips on height, for versatility.

landscape_size¶

(118mm 90mm)

Specify landscape page size (width and height). The program use this value when orientation is landscape.

toc_depth¶: (3)

[INT]

Specify (max) tree level of pdf bookmarks (Table of Contents). It uses html headings for structuring, so valid values are 0 to 6.

font_family¶

("DejaVu Sans", sans-serif)

Specify default font to use.

font_mono¶

("Dejavu Sans Mono", monospace)

Specify default monospaced font to use.

font_serif¶

(None)

Not used.

font_sans¶

(None)

Not used.

font_size¶

(9px)

Specify default font size.

font_size_mono¶

(8px)

Specify default monospaced font size.

font_scale¶

(1.0)

Specify scaling factor for css font_size and font_size_mono.

It is to make easier to test font sizes.

line_height¶

(1.3)

Specify default line height.

Converter Sections¶

Section prince and weasyprint are converter sections. They have common options.

When convert, only one converter is active, and only the options of that converter’s section are active.

commandline has the same options, to override.

Note

To see the current values for each converter:

$ tosixinch -a --prince
$ tosixinch -a --weasyprint

cnvpath¶

(prince)

The name or full path of the command as you type it in the shell. For ordinary installed ones, only the name would suffice.

css2¶

(None)

[COMMA]

Extra css files just to pass to converter commandline options.

It may be useful for converter specific features or troubles. Although, normally, you can do that better with css option and the template.

You can only use the filenames (not full paths).

The filenames are searched in css directory and current directory in order.

cnvopts¶: (None)

[CMD]

Additional options to pass to the command, besides css file option (which is added by css2 option above if it is specified).

site.ini¶

site.ini should have many sections, each is the settings for some specific site or a part of the site.

They all have the same options, in which the common options (the same ones as in tosixinch.ini) are not described here.

Each section must have match option. It is this option that is used as glob string to match input rsrcs, and consequently select which section to use.

So section names themselves can be arbitrary.

match¶

(None)

Glob string to match against input rsrc.

Path separator ('/') is not special for wildcards (*?[]!). So, e.g. '*' matches any strings including all subdirectories. (Actually, it uses fnmatch module, not glob module).

The program tries the values of this option from all the sections. The section whose match option matches the rsrc is used for the settings.

If there are multiple matches, the one with the most path separator characters ('/') is used (scheme separator '//' in 'https?://' are not counted). If there are multiple matches still, the last one is used.

If there is no match, default settings are used, and guess option is tried. In this case, a placeholder value http://tosixinch.example.com is set. (This imaginary site is used to make file paths in download and extract).

select¶

(None)

[LINE]

XPath strings to select elements from dfile when extract. Only selected elements are included in the <body> tag of the new efile, discarding others.

Each line in the value will be connected with a bar string ('|') when evaluating.

exclude¶: (None)

[LINE]

XPath strings to remove elements from the new efile after select. So you don’t need to exclude already excluded elements by select. As in select, each line in the value will be connected with a bar string ('|').

process¶

(None)

[LINE]

After select and exclude, arbitrary functions can be called if this option is specified.

Selection:

The functions must be top level ones.

It is searched in user process directory and application process directory, in order.

If the function name is found in multiple modules in user process directory, the program raises Error.

In that case, you can use dot notation. If the function name includes one dot ('.'), the program interprets it as <module name>.<function name>. Two or more dots are not supported.

Invocation:

The first argument of the functions is always doc, which the program provides. It is lxml.html DOM object (HtmlElement), corresponding to the resultant efile after select and exclude.

The function can have additional arguments. String after '?' (and before next '?') is interpreted as an argument.

For example, 'aaa?bb?cc' is made into code

if 'aaa.py' is found in user process directory:

process.aaa(doc, bb, cc)

or if it is found in application process directory:

tosixinch.process.aaa(doc, bb, cc)

You don’t have to return anything, just manipulate doc as you like. The program uses the resultant doc subsequently.

See process.sample for included sample functions.

Example:

Let’s say you want to change h3 tag to div for http://somesite.com.

First, create a file in process directory e.g. ~/.config/tosixinch/process/myprocess.py.

Second, create a top level function e.g.

def heading_to_div(doc, heading):
    """Change some heading to div from argument e.g. 'h3'."""
    for el in doc.xpath('//' + heading):
        el.tag = 'div'

Third, write configuration accordingly.

[somesite]
match=      http://somesite.com/*
select=     ...
process=    myprocess.heading_to_div?h3

cookie¶

(None)

[LINE]

Some sites require confirmation before providing the documents. (‘Are you over 18?’, ‘Agree to terms of service?’)

And urllib cannot handle these interactive communications.

By adding cookie data here (e.g. from your browsers), you may be able to bypass them.

Note it is not secure. Do not provide sensitive data.

dprocess¶

(None)

[LINE]

When download, the program runs functions specified by this option after getting http response, and before serializing to html text.

For completeness, it also runs when downloader is urllib, but the supposed usage is for other headless browsers.

For example, some webpages have folded contents which users need to click and run javascript to expand.

The mechanism is similar to process, Users define a function in a python file in user dprocess directory, with agent as the first argument, and modify it. If necessary, they can define other arguments by using '?' (see process).

But what comes as agent is dependent on what is actually downloader now:

urllib      http.client.HTTPResponse
selenium    selenium.webdriver.remote.webdriver.WebDriver

So user should be careful. (For example, when you define dprocess in site.ini, it is advisable to also define downloader).

Example:

def sitefoo_click(agent):  # for selenium
    path = '//div[@class="see_details"]'
    elements = agent.find_elements_by_xpath(path)
    for element in elements:
        element.click()
        time.sleep(1)

inspect¶

(get_links)

[LINE]

When action is inspect, the program runs functions this option specifies.

This is similar to extract action’s process, but inspect does not do anything before and after (select, exclude …, write to file).

Create Python functions in the same folder as process, original non-extracted html object is provided, as the first argument doc, and user do something, mostly print something.

See process.inspect_sample for a few sample functions.

Config Options¶

tosixinch.ini¶

General Section¶

Style Section¶

Converter Sections¶

site.ini¶

tosixinch

Navigation

Related Topics