API Documentation for the ebook editing tools

The ebook editing tools consist of a calibre.ebooks.oeb.polish.container.Container object that represents a book as a collection of HTML + resource files, and various tools that can be used to perform operations on the container. All the tools are in the form of module level functions in the various calibre.ebooks.oeb.polish.* modules.

You obtain a container object for a book at a path like this:

from calibre.ebooks.oeb.polish.container import get_container
container = get_container('Path to book file', tweak_mode=True)

If you are writing a plugin for the ebook editor, you get the current container for the book being edited like this:

from calibre.gui2.tweak_book import current_container
container = current_container()
if container is None:
    report_error # No book has been opened yet

The Container object

class calibre.ebooks.oeb.polish.container.Container(rootpath, opfpath, log, clone_data=None)[source]

A container represents an Open EBook as a directory full of files and an opf file. There are two important concepts:

  • The root directory. This is the base of the ebook. All the ebooks files are inside this directory or in its sub-directories.
  • Names: These are paths to the books’ files relative to the root directory. They always contain POSIX separators and are unquoted. They can be thought of as canonical identifiers for files in the book. Most methods on the container object work with names. Names are always in the NFC unicode normal form.
  • Clones: the container object supports efficient on-disk cloning, which is used to implement checkpoints in the ebook editor. In order to make this work, you should never access files on the filesystem directly. Instead, use raw_data() or open() to read/write to component files in the book.

When converting between hrefs and names use the methods provided by this class, they assume all hrefs are quoted.

abspath_to_name(fullpath, root=None)[source]

Convert an absolute path to a canonical name relative to root

Parameters:root – The base directory. By default the root for this container object is used.
add_file(name, data, media_type=None, spine_index=None, modify_name_if_needed=False, process_manifest_item=None)[source]

Add a file to this container. Entries for the file are automatically created in the OPF manifest and spine (if the file is a text document)

add_name_to_manifest(name, process_manifest_item=None)[source]

Add an entry to the manifest for a file with the specified name. Returns the manifest id.

apply_unique_properties(name, *properties)[source]

Ensure that the specified properties are set on only the manifest item identified by name. You can pass None as the name to remove the property from all items.

book_type = u'oeb'

The type of book (epub for EPUB files and azw3 for AZW3 files)

commit(outpath=None, keep_parsed=False)[source]

Commit all dirtied parsed objects to the filesystem and write out the ebook file at outpath.

  • output – The path to write the saved ebook file to. If None, the path of the original book file is used.
  • keep_parsed – If True the parsed representations of committed items are kept in the cache.
commit_item(name, keep_parsed=False)[source]

Commit a parsed object to disk (it is serialized and written to the underlying file). If keep_parsed is True the parsed representation is retained in the cache. See also: parsed()


Mark the parsed object corresponding to name as dirty. See also: parsed().


True iff a file/directory corresponding to the canonical name exists. Note that this function suffers from the limitations of the underlying OS filesystem, in particular case (in)sensitivity. So on a case insensitive filesystem this will return True even if the case of name is different from the case of the underlying filesystem file. See also has_name()


Return the size in bytes of the file represented by the specified canonical name. Automatically handles dirtied parsed objects. See also: parsed()

generate_item(name, id_prefix=None, media_type=None, unique_href=True)[source]

Add an item to the manifest with href derived from the given name. Ensures uniqueness of href and id automatically. Returns generated item.

get_file_path_for_processing(name, allow_modification=True)[source]

Similar to open() except that it returns a file path, instead of an open file object.


Mapping of guide type to canonical name


Return True iff a file with the same canonical name as that specified exists. Unlike exists() this method is always case-sensitive.

href_to_name(href, base=None)[source]

Convert an href (relative to base) to a name. base must be a name or None, in which case self.root is used.

insert_into_xml(parent, item, index=None)[source]

Insert item into parent (or append if index is None), fixing indentation. Only works with self closing items.

is_dir = False

If this container represents an unzipped book (a directory)

Iterate over all links in name. If get_line_numbers is True the yields results of the form (link, line_number, offset). Where line_number is the line_number at which the link occurs and offset is the number of characters from the start of the line. Note that offset could actually encompass several lines if not zero.


Ensure that name does not already exist in this book. If it does, return a modified version that does not exist.


Return True if the manifest has an entry corresponding to name


Mapping of manifest id to canonical names


The names of all manifest items whose media-type matches predicate. predicate can be a set, a list, a string or a function taking a single argument, which will be called with the media-type.


All manifest items that have the specified property


Mapping of manifest media-type to list of canonical names of that media-type


The metadata of this book as a Metadata object. Note that this object is constructed on the fly every time this property is requested, so use it sparingly.


Convert a canonical name to an absolute OS dependant path

name_to_href(name, base=None)[source]

Convert a name to a href relative to base, which must be a name or None in which case self.root is used as the base


Set of names that must never be renamed. Depends on the ebook file format.


Set of names that must never be deleted from the container. Depends on the ebook file format.


Set of names that are allowed to be missing from the manifest. Depends on the ebook file format.

open(name, mode=u'rb')[source]

Open the file pointed to by name for direct read/write. Note that this will commit the file if it is dirtied and remove it from the parse cache. You must finish with this file before accessing the parsed version of it again, or bad things will happen.


The parsed OPF file


Convenience method to either return the first XML element with the specified name or create it under the opf:package element and then return it, if it does not already exist.


The version set on the OPF’s <package> element


The version set on the OPF’s <package> element as a tuple of integers


Convenience method to evaluate an XPath expression on the OPF file, has the opf: and dc: namespace prefixes pre-defined.


Return a parsed representation of the file specified by name. For HTML and XML files an lxml tree is returned. For CSS files a cssutils stylesheet is returned. Note that parsed objects are cached for performance. If you make any changes to the parsed object, you must call dirty() so that the container knows to update the cache. See also replace().

raw_data(name, decode=True, normalize_to_nfc=True)[source]

Return the raw data corresponding to the file specified by name

  • decode – If True and the file has a text based mimetype, decode it and return a unicode object instead of raw bytes.
  • normalize_to_nfc – If True the returned unicode object is normalized to the NFC normal form as is required for the EPUB and AZW3 file formats.
relpath(path, base=None)[source]

Convert an absolute path (with os separators) to a path relative to base (defaults to self.root). The relative path is not a name. Use abspath_to_name() for that.

remove_from_spine(spine_items, remove_if_no_longer_in_spine=True)[source]

Remove the specified items (by canonical name) from the spine. If remove_if_no_longer_in_spine is True, the items are also deleted from the book, not just from the spine.


Removes item from parent, fixing indentation (works only with self closing items)

remove_item(name, remove_from_guide=True)[source]

Remove the item identified by name from this container. This removes all references to the item in the OPF manifest, guide and spine as well as from any internal caches.

rename(current_name, new_name)[source]

Renames a file from current_name to new_name. It automatically rebases all links inside the file if the directory the file is in changes. Note however, that links are not updated in the other files that could reference this file. This is for performance, such updates should be done once, in bulk.

replace(name, obj)[source]

Replace the parsed object corresponding to name with obj, which must be a similar object, i.e. an lxml tree for HTML/XML or a cssutils stylesheet for a CSS file.

Replace all links in name using replace_func, which must be a callable that accepts a URL and returns the replaced URL. It must also have a ‘replaced’ attribute that is set to True if any actual replacement is done. Convenient ways of creating such callables are using the LinkReplacer and LinkRebaser classes.


Convert a parsed object (identified by canonical name) into a bytestring. See parsed().


Set the spine to be spine_items where spine_items is an iterable of the form (name, linear). Will raise an error if one of the names is not present in the manifest.


An iterator yielding the path for every item in the books’ spine. See also: spine_iter and spine_items.


An iterator that yields item, name is_linear for every item in the books’ spine. item is the lxml element, name is the canonical file name and is_linear is True if the item is linear. See also: spine_names and spine_items.


An iterator yielding name and is_linear for every item in the books’ spine. See also: spine_iter and spine_items.

Managing component files in a container

Replace links to files in the container. Will iterate over all files in the container and change the specified links in them.

  • link_map – A mapping of old canonical name to new canonical name. For example: {'images/old.png': 'images/new.png'}
  • frag_map – A callable that takes two arguments (name, anchor) and returns a new anchor. This is useful if you need to change the anchors in HTML files. By default, it does nothing.
  • replace_in_opf – If False, links are not replaced in the OPF file.
calibre.ebooks.oeb.polish.replace.rename_files(container, file_map)[source]

Rename files in the container, automatically updating all links to them.

Parameters:file_map – A mapping of old canonical name to new canonical name, for example: {'text/chapter1.html': 'chapter1.html'}.

Return the folders that are recommended for the given filenames. The recommendation is based on where the majority of files of the same type are located in the container. If no files of a particular type are present, the recommended folder is assumed to be the folder containing the OPF file.

Pretty printing and auto fixing parse errors

calibre.ebooks.oeb.polish.pretty.fix_html(container, raw)[source]

Fix any parsing errors in the HTML represented as a string in raw. Fixing is done using the HTML5 parsing algorithm.


Fix any parsing errors in all HTML files in the container. Fixing is done using the HTML5 parsing algorithm.

calibre.ebooks.oeb.polish.pretty.pretty_html(container, name, raw)[source]

Pretty print the HTML represented as a string in raw

calibre.ebooks.oeb.polish.pretty.pretty_css(container, name, raw)[source]

Pretty print the CSS represented as a string in raw

calibre.ebooks.oeb.polish.pretty.pretty_xml(container, name, raw)[source]

Pretty print the XML represented as a string in raw. If name is the name of the OPF, extra OPF-specific prettying is performed.


Pretty print all HTML/CSS/XML files in the container

Managing book jackets


Remove an existing jacket, if any. Returns False if no existing jacket was found.


Either create a new jacket from the book’s metadata or replace an existing jacket. Returns True if an existing jacket was replaced.

Splitting and merging of files

calibre.ebooks.oeb.polish.split.split(container, name, loc_or_xpath, before=True, totals=None)[source]

Split the file specified by name at the position specified by loc_or_xpath. Splitting automatically migrates all links and references to the affected files.

  • loc_or_xpath – Should be an XPath expression such as //h:div[@id=”split_here”]. Can also be a loc which is used internally to implement splitting in the preview panel.
  • before – If True the split occurs before the identified element otherwise after it.
  • totals – Used internally
calibre.ebooks.oeb.polish.split.multisplit(container, name, xpath, before=True)[source]

Split the specified file at multiple locations (all tags that match the specified XPath expression). See also: split(). Splitting automatically migrates all links and references to the affected files.

Parameters:before – If True the splits occur before the identified element otherwise after it.
calibre.ebooks.oeb.polish.split.merge(container, category, names, master)[source]

Merge the specified files into a single file, automatically migrating all links and references to the affected files. The file must all either be HTML or CSS files.

  • category – Must be either 'text' for HTML files or 'styles' for CSS files
  • names – The list of files to be merged
  • master – Which of the merged files is the master file, that is, the file that will remain after merging.

Managing covers

calibre.ebooks.oeb.polish.cover.set_cover(container, cover_path, report=None, options=None)[source]

Set the cover of the book to the image pointed to by cover_path.

  • cover_path – Either the absolute path to an image file or the canonical name of an image in the book. When using an image in the book, you must also set options, see below.
  • report – An optional callable that takes a single argument. It will be called with information about the tasks being processed.
  • options – None or a dictionary that controls how the cover is set. The dictionary can have entries: keep_aspect: True or False (Preserve aspect ratio of covers in EPUB) no_svg: True or False (Use an SVG cover wrapper in the EPUB titlepage) existing: True or False (cover_path refers to an existing image in the book)
calibre.ebooks.oeb.polish.cover.mark_as_cover(container, name)[source]

Mark the specified image as the cover image.

calibre.ebooks.oeb.polish.cover.mark_as_titlepage(container, name, move_to_start=True)[source]

Mark the specified HTML file as the titlepage of the EPUB.

Parameters:move_to_start – If True the HTML file is moved to the start of the spine

Working with CSS

calibre.ebooks.oeb.polish.fonts.change_font(container, old_name, new_name=None)[source]

Change a font family from old_name to new_name. Changes all occurrences of the font family in stylesheets, style tags and style attributes. If the old_name refers to an embedded font, it is removed. You can set new_name to None to remove the font family instead of changing it.

calibre.ebooks.oeb.polish.css.remove_unused_css(container, report=None, remove_unused_classes=False, merge_rules=False)[source]

Remove all unused CSS rules from the book. An unused CSS rule is one that does not match any actual content.

  • report – An optional callable that takes a single argument. It is called with information about the operations being performed.
  • remove_unused_classes – If True, class attributes in the HTML that do not match any CSS rules are also removed.
  • merge_rules – If True, rules with identical selectors are merged.
calibre.ebooks.oeb.polish.css.filter_css(container, properties, names=())[source]

Remove the specified CSS properties from all CSS rules in the book.

  • properties – Set of properties to remove. For example: {'font-family', 'color'}.
  • names – The files from which to remove the properties. Defaults to all HTML and CSS files in the book.

Working with the Table of Contents

calibre.ebooks.oeb.polish.toc.from_xpaths(container, xpaths)[source]

Generate a Table of Contents from a list of XPath expressions. Each expression in the list corresponds to a level of the generate ToC. For example: ['//h:h1', '//h:h2', '//h:h3'] will generate a three level table of contents from the <h1>, <h2> and <h3> tags.

Generate a Table of Contents from links in the book.


Generate a Table of Contents from files in the book.

calibre.ebooks.oeb.polish.toc.create_inline_toc(container, title=None)[source]

Create an inline (HTML) Table of Contents from an existing NCX table of contents.

Parameters:title – The title for this table of contents.

Edit Book Tool

class calibre.gui2.tweak_book.plugin.Tool[source]

Bases: object

The base class for individual tools in an Edit Book plugin. Useful members include:

Methods that must be overridden in sub classes:

name = None

Set this to a unique name it will be used as a key

allowed_in_toolbar = True

If True the user can choose to place this tool in the plugins toolbar

allowed_in_menu = True

If True the user can choose to place this tool in the plugins menu

toolbar_button_popup_mode = u'delayed'

The popup mode for the menu (if any) of the toolbar button. Possible values are ‘delayed’, ‘instant’, ‘button’


The calibre.gui2.tweak_book.boss.Boss object. Used to control the user interface.


The main window of the user interface


Return the current calibre.ebooks.oeb.polish.container.Container object that represents the book being edited.

register_shortcut(qaction, unique_name, default_keys=(), short_text=None, description=None, **extra_data)[source]

Register a keyboard shortcut that will trigger the specified qaction. This keyboard shortcut will become automatically customizable by the user in the Keyboard section of the editor preferences.

  • qaction – A QAction object, it will be triggered when the configured key combination is pressed by the user.
  • unique_name – A unique name for this shortcut/action. It will be used internally, it must not be shared by any other actions in this plugin.
  • default_keys – A list of the default keyboard shortcuts. If not specified no default shortcuts will be set. If the shortcuts specified here conflict with either builtin shortcuts or shortcuts from user configuration/other plugins, they will be ignored. In that case, users will have to configure the shortcuts manually via Preferences. For example: default_keys=('Ctrl+J', 'F9').
  • short_text – An optional short description of this action. If not specified the text from the QAction will be used.
  • description – An optional longer description of this action, it will be used in the preferences entry for this shortcut.

Create a QAction that will be added to either the plugins toolbar or the plugins menu depending on for_toolbar. For example:

def create_action(self, for_toolbar=True):
    ac = QAction(get_icons('myicon.png'), 'Do something')
    if for_toolbar:
        # We want the toolbar button to have a popup menu
        menu = QMenu()
        menu.addAction('Do something else')
        subaction = menu.addAction('And another')

        # Register a keyboard shortcut for this toolbar action be
        # careful to do this for only one of the toolbar action or
        # the menu action, not both.
        self.register_shortcut(ac, 'some-unique-name', default_keys=('Ctrl+K',))
    return ac

See also

Method register_shortcut().

Controlling the editor’s user interface

The ebook editor’s user interface is controlled by a single global Boss object. This has many useful methods that can be used in plugin code to perform common tasks.

class calibre.gui2.tweak_book.boss.Boss(parent, notify=None)[source]

Create a restore checkpoint with the name specified as msg


Update all the components of the user interface to reflect the latest data in the current book container.

Parameters:mark_as_modified – If True, the book will be marked as modified, so the user will be prompted to save it when quitting.

Close the editor that is editing the file specified by name


Commit any changes that the user has made to files open in editors to the container. You should call this method before performing any actions on the current container


Return the name of the file being edited currently or None if no file is being edited

edit_file(name, syntax=None, use_template=None)[source]

Open the file specified by name in an editor

  • syntax – The media type of the file, for example, 'text/html'. If not specified it is guessed from the file extension.
  • use_template – A template to initialize the opened editor with
open_book(path=None, edit_file=None, clear_notify_data=True, open_folder=False)[source]

Open the ebook at path for editing. Will show an error if the ebook is not in a supported format or the current book has unsaved changes.

Parameters:edit_file – The name of a file inside the newly opened book to start editing. Can also be a list of names.

Undo the previous creation of a restore checkpoint, useful if you create a checkpoint, then abort the operation with no changes


Save the book. Saving is performed in the background


Mark the book as having been modified

show_current_diff(allow_revert=True, to_container=None)[source]

Show the changes to the book from its last checkpointed state

  • allow_revert – If True the diff dialog will have a button to allow the user to revert all changes
  • to_container – A container object to compare the current container to. If None, the previously checkpointed container is used

Show the editor that is editing the file specified by name


Sync the position of the preview panel to the current cursor position in the current editor