Refactor PR to use pre-parsed SMIL metadata

src/calibre/srv/render_book.py

@@ -354,9 +354,12 @@ def transform_smil(container, name, link_uid, virtualize_resources, virtualized_
     text_tag, audio_tag = SMIL('text'), SMIL('audio')
     body_tag, seq_tag, par_tag = SMIL('body'), SMIL('seq'), SMIL('par')
     type_attr, textref_attr = EPUB('type'), EPUB('textref')
+    parnum = 0
 
-    def make_par(par):
-        ans = {}
+    def make_par(par, target):
+        nonlocal parnum
+        parnum += 1
+        ans = {'num': parnum}
         t = par.get(type_attr)
         if t:
             ans['type'] = t
@@ -364,7 +367,10 @@ def transform_smil(container, name, link_uid, virtualize_resources, virtualized_
             if child.tag == text_tag:
                 src = child.get('src')
                 if src:
-                    ans['text'] = [container.href_to_name(child.get('src'), name), src.partition('#')[2]]
+                    q = container.href_to_name(src, name)
+                    if q != target:
+                        return {}  # the par must match the textref of the parent seq
+                    ans['anchor'] = src.partition('#')[2]
             elif child.tag == audio_tag:
                 src = child.get('src')
                 if src:
@@ -386,10 +392,15 @@ def transform_smil(container, name, link_uid, virtualize_resources, virtualized_
             parent_seq = smil_map.get(target)
             if parent_seq is None:
                 smil_map[target] = parent_seq = {'textref': [target, ''], 'par':[], 'seq':[], 'type': 'root'}
+        else:
+            if parent_seq['textref'][0] != target:
+                return  # child seqs must be in the same HTML file as parent
         parent_seq['seq'].append(seq)
         for child in seq_xml_element.iterchildren('*'):
             if child.tag == par_tag:
-                seq['par'].append(make_par(child))
+                p = make_par(child, target)
+                if p.get('audio'):
+                    seq['par'].append(p)
             elif child.tag == seq_tag:
                 tref = child.get(textref_attr)
                 if tref:

src/pyj/read_book/flow_mode.pyj

@@ -597,6 +597,9 @@ anchor_funcs = {
     'cmp': def cmp(a, b):
         return (a.block - b.block) or (a.inline - b.inline)
     ,
+    'get_bounding_client_rect': def(elem):
+        return elem.getBoundingClientRect()
+    ,
 }
 
 

src/pyj/read_book/iframe.pyj

@@ -60,6 +60,7 @@ from read_book.settings import (
 from read_book.shortcuts import (
     create_shortcut_map, keyevent_as_shortcut, shortcut_for_key_event
 )
+from read_book.smil import flatten_smil_map, smil_element_at, mark_smil_element
 from read_book.toc import find_anchor_before_range, update_visible_toc_anchors
 from read_book.touch import (
     create_handlers as create_touch_handlers, reset_handlers as reset_touch_handlers
@@ -67,7 +68,7 @@ from read_book.touch import (
 from read_book.viewport import scroll_viewport
 from select import (
     first_visible_word, is_start_closer_to_point, move_end_of_selection,
-    selection_extents, word_at_point, span_id_at_point, id_of_first_visible_span
+    selection_extents, word_at_point
 )
 from utils import debounce, is_ios
 
@@ -259,6 +260,8 @@ class IframeBoss:
         set_toc_anchor_map()
         self.replace_history_on_next_cfi_update = True
         self.book = current_book.book = data.book
+        self.smil_map = data.smil_map
+        self.smil_anchor_map, self.smil_par_list = flatten_smil_map(self.smil_map)
         self.link_attr = 'data-' + self.book.manifest.link_uid
         self.reference_mode_enabled = data.reference_mode_enabled
         self.is_titlepage = data.is_titlepage
@@ -475,6 +478,8 @@ class IframeBoss:
                 window.setTimeout(def():
                     self.annotations_msg_received({'type': 'edit-highlight', 'uuid': ipos.uuid})
                 , 5)
+            elif ipos.type is 'smil_id':
+                self.audio_ebook_msg_received({'type': 'play', 'anchor': ipos.anchor})
         spine = self.book.manifest.spine
         files = self.book.manifest.files
         spine_index = csi.index
@@ -1009,35 +1014,30 @@ class IframeBoss:
 
     def audio_ebook_msg_received(self, data):
         if data.type is 'mark':
-            self.color_span_id(data.old_span_id, data.span_id)
-        elif data.type is 'play':
-            if data.pos:
-                span_id = span_id_at_point(data.pos.x, data.pos.y)
-                self.send_message('audio_ebook_message', type='report-span-id', span_id=span_id)
+            if data.anchor:
+                self.last_search_at = window.performance.now()
+                if mark_smil_element(data.anchor):
+                    self.ensure_selection_boundary_visible()
+                    self.send_message('audio_ebook_message', type='marked', anchor=data.anchor, idx=data.idx)
+                else:
+                    self.send_message('audio_ebook_message', type='marked')
             else:
-                span_id = id_of_first_visible_span()
-                self.send_message('audio_ebook_message', type='report-span-id', span_id=span_id)
-
-        elif data.type is 'trigger-shortcut':
+                window.getSelection().removeAllRanges()
+                self.send_message('audio_ebook_message', type='marked')
+        elif data.type is 'play':
+            if data.anchor:
+                pos = self.smil_anchor_map[data.anchor]
+                if pos?:
+                    par = self.smil_par_list[pos]
+                else:
+                    par = None
+            else:
+                par = smil_element_at(data.pos, self.smil_anchor_map, self.smil_par_list)
+            self.send_message('audio_ebook_message', type='start_play_at', par=par or None, anchor=data.anchor or None)
+        elif data.type is 'trigger_shortcut':
             self.on_handle_navigation_shortcut(data)
-
-    def color_span_id(self, old_span_id, span_id):
-        def element_in_viewport(element):
-            rect = element.getBoundingClientRect()
-            return (
-                rect.top >= 0 and
-                rect.left >= 0 and
-                rect.bottom <= (window.innerHeight or document.documentElement.clientHeight) and
-                rect.right <= (window.innerWidth or document.documentElement.clientWidth)
-            )
-        element = document.getElementById(span_id)
-        old_element = document.getElementById(old_span_id)
-        if old_element:
-            old_element.style.backgroundColor = ''
-        if element:
-            element.style.backgroundColor = window.getComputedStyle(document.documentElement, '::selection').backgroundColor
-            if not element_in_viewport(element):
-                scroll_to_elem(element)
+        else:
+            console.error(f'Unknown audio ebook message type from main: {data.type}')
 
     def hints_msg_received(self, data):
         if data.type is 'show':

src/pyj/read_book/paged_mode.pyj

@@ -879,6 +879,7 @@ anchor_funcs = {
     'cmp': def cmp(a, b):
         return a - b
     ,
+    'get_bounding_client_rect': get_bounding_client_rect,
 }
 
 

src/pyj/read_book/read_aloud.pyj

@@ -8,7 +8,7 @@ from book_list.globals import get_session_data
 from book_list.theme import get_color
 from dom import clear, svgicon, unique_id
 from gettext import gettext as _
-from read_book.globals import runtime, ui_operations
+from read_book.globals import ui_operations
 from read_book.highlights import ICON_SIZE
 from read_book.selection_bar import BUTTON_MARGIN, get_margins, map_to_iframe_coords
 from read_book.shortcuts import shortcut_for_key_event
@@ -58,7 +58,7 @@ class ReadAloud:
 
     @property
     def supports_css_min_max(self):
-        return not runtime.is_standalone_viewer or runtime.QT_VERSION >= 0x050f00
+        return True
 
     @property
     def bar(self):

src/pyj/read_book/read_audio_ebook.pyj

@@ -2,32 +2,28 @@
 # License: GPL v3 Copyright: 2023, DO LE DUY <duy.dole.00ece at gmail.com>
 
 
-# The key difference between an ePub with SMIL audio synchronization (EPUB3 with Media Overlays) and a regular ePub is the inclusion of SMIL files and audio content:
-
-# SMIL Files: ePub with SMIL includes SMIL (Synchronized Multimedia Integration Language) files, XML documents that define audio and text synchronization.
-
-# Audio Content: It contains audio files that match eBook sections, referenced in SMIL files for synchronized playback.
-
-# Text Content: The textual content, often in HTML or XHTML files, remains similar to regular ePub. Text and audio are linked using <span> tags with unique IDs.
-
-# SMIL, audio, and text files are organized into folders, usually inside the epub/ or OEBPS/ folder. Sometimes, SMIL files may be placed in text folders. In this program we assume that each spoken text file corresponds to one audio file and one SMIL file.  
-
-# Public domain audio eBooks can be found on https://www.readbeyond.it/ebooks.html. ReadBeyond also offers Aeneas (https://github.com/readbeyond/aeneas), an open-source tool for force-alignment of audio and text to generate smil files. Another notable tool is https://github.com/r4victor/syncabook, builds upon Aeneas to complete a workflow for creating EPUB3 with Media Overlays.
-
+# Public domain audio eBooks can be found on https://www.readbeyond.it/ebooks.html.
+# ReadBeyond also offers Aeneas (https://github.com/readbeyond/aeneas),
+# an open-source tool for force-alignment of audio and text to generate smil files.
+# Another notable tool is https://github.com/r4victor/syncabook,
+# builds upon Aeneas to complete a workflow for creating EPUB3 with Media Overlays.
 
 from __python__ import bound_methods, hash_literals
 
 from elementmaker import E
 
-from book_list.globals import get_session_data
-from dom import svgicon, unique_id, change_icon_image, clear
+from dom import change_icon_image, svgicon, unique_id
 from gettext import gettext as _
-from book_list.theme import get_color
-from read_book.globals import runtime, ui_operations, current_spine_item, current_book
+from modals import error_dialog
+from read_book.globals import ui_operations
 from read_book.highlights import ICON_SIZE
 from read_book.selection_bar import BUTTON_MARGIN, get_margins, map_to_iframe_coords
 from read_book.shortcuts import shortcut_for_key_event
-from modals import question_dialog
+from read_book.smil import (
+    find_next_audio_in_spine, get_smil_id_for_timestamp, next_audio_file_for_spine_item
+)
+
+HIDDEN, PLAYING, PAUSED = 'HIDDEN', 'PLAYING', 'PAUSED'
 
 
 class ReadAudioEbook:
@@ -36,12 +32,17 @@ class ReadAudioEbook:
 
     def __init__(self, view):
         self.view = view
-        self.parser = new window.DOMParser()
-        self._state = "HIDDEN"
+        self._state = HIDDEN
+        self.initialized = False
+        self.current_audio_src = ''
+
+    def initialize(self):
+        if self.initialized:
+            return
+        self.initialized = True
         self.bar_id = unique_id("bar")
         self.overlay_off = False
-        self.container.style.height = "100%"
-        
+
         container = self.container
         container.style.transition = "height 0.5s ease-in-out"
         container.style.backgroundColor = "rgba(127, 127, 127, 0.05)"
@@ -49,7 +50,7 @@ class ReadAudioEbook:
         container.appendChild(E.div(
             id=self.bar_id,
             style="position: absolute; bottom: 0; width: 90%; height: 2em; border-radius: 1em; padding:0.5em; display: flex; justify-content: center; align-items: center; background-color: rgba(127, 127, 127, 0.3); "
-        )) 
+        ))
         container.addEventListener("keydown", self.on_keydown, {"passive": False})
         container.addEventListener("click", self.on_container_clicked, {"passive": False})
         container.addEventListener("contextmenu", self.toggle, {"passive": False})
@@ -92,119 +93,112 @@ class ReadAudioEbook:
                 create_button("faster", "faster", _("Speed up audio")),
                 create_button("hide", "off", _("Close Read Audio-Ebook"))
             )
-            
+
         ]:
             bar_container.appendChild(x)
 
-        self.audio_buttons = document.getElementById("audioButtons") 
-        self.progress_bar = document.getElementById("progressBar") 
+        self.audio_buttons = document.getElementById("audioButtons")
+        self.progress_bar = document.getElementById("progressBar")
         self.time_display = document.getElementById("timeDisplay")
 
-        self.audio_id = unique_id("audio")
-        self.container.appendChild(E.audio(
-            id=self.audio_id,
-            style="display:none"
-        )) 
+        self.container.appendChild(E.audio(id=unique_id('audio'), style="display:none"))
+        ap = self.container.lastChild
+        self.audio_player_id = ap.id
 
-        self.audio_player = document.getElementById(self.audio_id) 
-
-        self.audio_player.addEventListener("timeupdate", def():
-            if self.state != "HIDDEN":
-                if self.audio_player.duration:
-                    audio_current_time = self.audio_player.currentTime
-
-                    progress = (audio_current_time / self.audio_player.duration) * 100
+        ap.addEventListener("timeupdate", def():
+            if self.state is not HIDDEN:
+                ap = self.audio_player
+                if ap.duration:
+                    audio_current_time = ap.currentTime
+                    progress = (audio_current_time / ap.duration) * 100
                     self.progress_bar.firstChild.style.width = progress + "%"
-
-                    self.time_display.textContent = f"{self.seconds_to_ms(audio_current_time)}/{self.seconds_to_ms(self.audio_player.duration)}"
-
-                    span_id = self.find_span_id_for_time(audio_current_time)
-                    if span_id != self.span_id:
-                        old_span_id = self.span_id
-                        self.span_id = span_id
-                        self.send_message("mark", old_span_id=old_span_id, span_id = self.span_id)
+                    self.time_display.textContent = f"{self.seconds_to_ms(audio_current_time)}/{self.seconds_to_ms(ap.duration)}"
+                    self.mark_for_timeupdate(audio_current_time)
                 else:
                     self.time_display.textContent = "00:00"
                     self.progress_bar.firstChild.style.width = "0%"
 
         )
 
-        self.audio_player.addEventListener("ended", def():
-            self.view.show_next_spine_item()
-        )  
+        ap.addEventListener("ended", def():
+            self.play_next_audio_file()
+        )
 
         self.progress_bar.addEventListener("click", def(event):
-            if self.audio_player.duration:
+            ap = self.audio_player
+            if ap.duration:
                 rect = self.progress_bar.getBoundingClientRect()
                 clickX = event.clientX - rect.left
                 total_width = rect.width
-                skip_time = (clickX / total_width) * self.audio_player.duration
-                self.audio_player.currentTime = skip_time
+                skip_time = (clickX / total_width) * ap.duration
+                ap.currentTime = skip_time
         )
 
+    def mark_for_timeupdate(self, audio_time):
+        sam = self.view.currently_showing.smil_audio_map
+        if sam:
+            smil_id, idx = get_smil_id_for_timestamp(self.current_audio_src, audio_time, sam, self.last_marked_smil_idx)
+            if smil_id and smil_id is not self.last_marked_smil_id:
+                self.send_message('mark', anchor=smil_id, idx=idx)
 
-    def parse_smil_file(self, smil_content, smil_name, mimetype):
-        # Extract information from the parsed XML
-        smil_content.text().then(def(data): 
-            xml_doc = self.parser.parseFromString(data, "text/xml")
-            audio_map = {}
-            par_elements = xml_doc.getElementsByTagName("par")
-
-            audio_element = par_elements[0].getElementsByTagName("audio")[0]
-            audio_file = audio_element.getAttribute("src")
-
-            for par_element in par_elements:
-                text_element = par_element.getElementsByTagName("text")[0]
-                audio_element = par_element.getElementsByTagName("audio")[0]
-                if text_element and audio_element:
-                    span_id = text_element.getAttribute("src").split("#")[1]
-                    audio_details = {
-                        "clipBegin": audio_element.getAttribute("clipBegin"),
-                        "clipEnd": audio_element.getAttribute("clipEnd")
-                    }
-                    audio_map[span_id] = audio_details
-
-            self.audio_maps[smil_name[:-5].replace("smil", "text")] = [audio_map, audio_file]
-        )
-
-    def change_audio_src(self):
-        window.URL.revokeObjectURL(self.audio_player.src)
-        self.audio_player.setAttribute("src", "")
-        if self.audio_maps[current_spine_item()]:
-            self.audio_map = self.audio_maps[current_spine_item()]
-            link = self.audio_files[0].split("/")[0] + self.audio_maps[current_spine_item()][1][2:]
-            ui_operations.get_file(
-                current_book(), link, def(blob, name, mimetype):
-                    self.pause()
-                    blob_url = window.URL.createObjectURL(blob)
-                    self.audio_player.src = blob_url
-                    self.send_message("play")
+    def play_next_audio_file(self):
+        next_audio_file, par = next_audio_file_for_spine_item(self.current_audio_src, self.view.currently_showing.smil_audio_map)
+        if next_audio_file:
+            self.set_audio_src(next_audio_file, def():
+                self.audio_player.currentTime = par.audio?.start or 0
+                self.play()
             )
-            # self.pause()
-            # self.audio_player.src = "book/" + link
-            # self.send_message("play")
-
+            return
+        spine_name, par = find_next_audio_in_spine(self.view.currently_showing.name, self.view.book.manifest)
+        if spine_name:
+            self.view.show_name(spine_name, initial_position={'type': 'smil_id', 'anchor': par.anchor})
         else:
-            if self.state is "PLAYING":
-                if self.skip_section:
-                    self.view.show_next_spine_item()
-                else:
-                    self.pause()
-                    question_dialog(_('Skip Section'), _('Do you want to automatically skip sections without audio?'), def (yes):
-                        if yes:
-                            self.skip_section = True
-                            self.play()
-                        else:
-                            self.hide()
-                    )
+            self.hide()
+
+    def set_audio_src(self, name, proceed):
+        if self.current_audio_src is name:
+            proceed()
+            return
+        self.last_marked_smil_id = self.last_marked_smil_idx = None
+        self.current_audio_src = name
+        ap = self.audio_player
+        if ui_operations.get_url_for_book_file_name:
+            ap.onloadeddata = def():
+                proceed()
+            ap.onerror = def(evt):
+                console.error(evt)
+                error_dialog(_('Could not load audio'), _(
+                    'Could not load the audio file: {} with error: {}').format(name, evt.message))
+            ap.src = ui_operations.get_url_for_book_file_name(name)
+            return
+        if ap.src:
+            ap.onloadeddata = def():
+                pass
+            ap.onerror = def():
+                pass
+            window.URL.revokeObjectURL(ap.src)
+            ap.src = ''
+        ui_operations.get_file(self.view.book, name, def(blob, name, mimetype):
+            ap = self.audio_player
+            ap.onloadeddata = def():
+                proceed()
+            ap.onerror = def(evt):
+                console.error(evt)
+                error_dialog(_('Could not load audio'), _(
+                    'Could not load the audio file: {} with error: {}').format(name, evt.message))
+            ap.src = window.URL.createObjectURL(blob)
+        )
+
+    def start_playback(self):
+        self.send_message('play')
 
     @property
     def container(self):
         return document.getElementById("audio-ebooks-overlay")
 
     @property
-    def supports_css_min_max(self):
-        return not runtime.is_standalone_viewer or runtime.QT_VERSION >= 0x050f00
+    def audio_player(self):
+        return document.getElementById(self.audio_player_id)
 
     @property
     def is_visible(self):
@@ -216,24 +210,30 @@ class ReadAudioEbook:
 
     @state.setter
     def state(self, val):
-        if val is not self._state:
-            self._state = val
+        if val is not self._state :
+            if val is HIDDEN:
+                self._state = HIDDEN
+            elif val is PLAYING:
+                self._state = PLAYING
+            elif val is PAUSED:
+                self._state = PAUSED
 
     def hide(self):
-        if self.state is not "HIDDEN":
-            self.send_message("mark", old_span_id=self.span_id)
+        if self.state is not HIDDEN:
+            self.send_message("mark")
             self.pause()
             self.container.style.display = "none"
             self.view.focus_iframe()
-            self.state = "HIDDEN"
+            self.state = HIDDEN
             if ui_operations.read_aloud_state_changed:
                 ui_operations.read_aloud_state_changed(False)
 
     def show(self):
-        if self.state is "HIDDEN":
-            self.state = "PLAYING"
+        if self.state is HIDDEN:
+            self.initialize()
+            self.state = PLAYING
             change_icon_image(document.getElementById("audio-ebook-bt-toggle"), "pause")
-            self.change_audio_src()
+            self.start_playback()
             self.container.style.display = "block"
             self.focus()
             if ui_operations.read_aloud_state_changed:
@@ -244,28 +244,28 @@ class ReadAudioEbook:
 
     def slower(self):
         self.audio_player.playbackRate -= 0.1
-        
+
     def faster(self):
         self.audio_player.playbackRate += 0.1
 
     def play(self):
-        self.state = "PLAYING" 
+        self.state = PLAYING
         change_icon_image(document.getElementById("audio-ebook-bt-toggle"), "pause")
-        if self.audio_player.getAttribute("src"):
-            self.audio_player.play()
-        else:
-            self.view.show_next_spine_item()
-        
+        ap = self.audio_player
+        if ap.getAttribute("src"):
+            ap.play()
+
     def pause(self):
-        self.state = "PAUSED"
+        self.state = PAUSED
         change_icon_image(document.getElementById("audio-ebook-bt-toggle"), "play")
-        if self.audio_player.getAttribute("src"):
-            self.audio_player.pause()
+        ap = self.audio_player
+        if ap.getAttribute("src"):
+            ap.pause()
 
     def toggle(self):
-        if self.state is "PLAYING":
+        if self.state is PLAYING:
             self.pause()
-        elif self.state is "PAUSED":
+        elif self.state is PAUSED:
             self.play()
 
     def overlay(self):
@@ -277,7 +277,7 @@ class ReadAudioEbook:
             self.overlay_off = True
             self.container.style.height = "3em"
             change_icon_image(document.getElementById("audio-ebook-bt-overlay"), "overlay-on")
-            
+
     def on_container_clicked(self, ev):
         if ev.button is not 0:
             return
@@ -309,23 +309,7 @@ class ReadAudioEbook:
         elif sc_name is "quit":
             self.hide()
         elif sc_name in ("up", "down", "pageup", "pagedown", "left", "right"):
-            self.send_message("trigger-shortcut", name=sc_name)
-    
-    def find_span_id_for_time(self, current_time):
-        if self.audio_map:
-            for span_id in self.audio_map[0]:
-                clip_begin_time = self.convert_time_to_seconds(self.audio_map[0][span_id]["clipBegin"])
-                clip_end_time = self.convert_time_to_seconds(self.audio_map[0][span_id]["clipEnd"])               
-                if clip_begin_time <= current_time < clip_end_time:
-                    return span_id
-        return None
-
-    def convert_time_to_seconds(self, time_string):
-        parts = time_string.split(":")
-        if len(parts) != 3:
-            return 0
-        hours, minutes, seconds = map(float, parts)
-        return hours * 3600 + minutes * 60 + seconds
+            self.send_message("trigger_shortcut", name=sc_name)
 
     def seconds_to_ms(self, seconds):
         minutes = Math.floor(seconds / 60)
@@ -336,10 +320,26 @@ class ReadAudioEbook:
         self.view.iframe_wrapper.send_message("audio-ebook", type=message_type, **kw)
 
     def handle_message(self, message):
-        if message.type is "report-span-id":
-            if message.span_id:
-                old_span_id = self.span_id
-                self.span_id = message.span_id
-                self.send_message("mark", old_span_id=old_span_id, span_id=self.span_id)
-                self.audio_player.currentTime = self.convert_time_to_seconds(self.audio_map[0][self.span_id]["clipBegin"])
-                self.play()
+        if message.type is 'start_play_at':
+            if message.par:
+                self.set_audio_src(
+                    message.par.audio, def():
+                        self.audio_player.currentTime = message.par.start or 0
+                        self.play()
+                )
+            else:
+                if message.anchor:
+                    # start playing from where we are
+                    self.send_message('play')
+                    return
+                self.pause()
+                error_dialog(_('Audio element not found'), _(
+                    'Could not play audio as no associated audio was found'))
+        elif message.type is 'marked':
+            if message.anchor:
+                self.last_marked_smil_id = message.anchor
+                self.last_marked_smil_idx = message.idx
+            else:
+                self.last_marked_smil_id = self.last_marked_smil_idx = None
+        else:
+            console.error(f'Unknown audio ebook message type from iframe: {message.type}')

src/pyj/read_book/smil.pyj

@@ -0,0 +1,132 @@
+# vim:fileencoding=utf-8
+# License: GPL v3 Copyright: 2023, Kovid Goyal <kovid at kovidgoyal.net>
+from __python__ import bound_methods, hash_literals
+
+from read_book.anchor_visibility import is_anchor_on_screen
+from read_book.globals import get_boss
+
+
+def flatten_seq(seq, par_list):
+    if seq.par:
+        for par in seq.par:
+            if par.anchor:
+                par_list.push(par)
+    if seq.seq:
+        for child in seq.seq:
+            flatten_seq(child, par_list)
+
+
+def flatten_smil_map(smil_map):
+    anchor_map = {}
+    par_list = v'[]'
+    if smil_map:
+        flatten_seq(smil_map, par_list)
+    par_list.sort(def (a, b): return a.num - b.num;)
+    for i in range(par_list.length):
+        anchor_map[par_list[i].anchor] = i
+    return anchor_map, par_list
+
+
+def get_smil_audio_map(smil_map):
+    audio_map = {}
+
+    def flatten(seq):
+        if seq.par:
+            for par in seq.par:
+                if par.audio:
+                    a = audio_map[par.audio]
+                    if not a:
+                        a = audio_map[par.audio] = v'[]'
+                    a.push(par)
+        if seq.seq:
+            for child in seq.seq:
+                flatten(child)
+    flatten(smil_map)
+    for v in Object.values(audio_map):
+        v.sort(def(a, b): return a.start - b.start;)
+    return {'audio_files': audio_map}
+
+
+def get_smil_id_for_timestamp(audio_file_name, timestamp, smil_audio_map, prev_idx):
+    pars = smil_audio_map.audio_files[audio_file_name]
+    if not pars:
+        return None, None
+    prev_idx = prev_idx or 0
+    if prev_idx >= pars.length or prev_idx < 0:
+        prev_idx = 0
+    for i in range(prev_idx, pars.length):
+        if pars[i].start <= timestamp <= pars[i].end:
+            return pars[i].anchor, i
+    for i in range(0, prev_idx):
+        if pars[i].start <= timestamp <= pars[i].end:
+            return pars[i].anchor, i
+    return None, None
+
+
+def next_audio_file_for_spine_item(audio_file_name, smil_audio_map):
+    if audio_file_name:
+        keys = Object.keys(smil_audio_map)
+        idx = keys.indexOf(audio_file_name)
+        if 0 <= idx < keys.length - 1:
+            return keys[idx+1], smil_audio_map[keys[idx+1]]
+    return None, None
+
+
+def first_par(smil_map):
+    par_list = flatten_smil_map(smil_map)[1]
+    for par in par_list:
+        if par.anchor:
+            return par
+    return None
+
+
+def find_next_audio_in_spine(current_name, book_manifest):
+    spine = book_manifest.spine
+    file_map = book_manifest.files
+    spine_idx = spine.indexOf(current_name)
+
+    for i in range(spine_idx + 1, spine.length):
+        q = spine[i]
+        f = file_map[q]
+        if f and f.smil_map:
+            par = first_par(f.smil_map)
+            if par:
+                return q, par
+    return None, None
+
+
+def smil_element_at(pos, anchor_map, par_list):
+    if pos:
+        # first see if we get lucky
+        elem = document.elementFromPoint(pos.x, pos.y)
+        if elem and elem.id and anchor_map[elem.id]?:
+            return par_list[anchor_map[elem.id]]
+        # now try to find a par that intersects pos
+        af = get_boss().anchor_funcs
+        for par in par_list:
+            if par.anchor and par.audio:
+                elem = document.getElementById(par.anchor)
+                if elem:
+                    br = af.get_bounding_client_rect(elem)
+                    if br.x <= pos.x <= (br.x + br.width) and br.y <= pos.y <= (br.y + br.height):
+                        return par
+        # fallback to first visible anchor
+        return smil_element_at(None, anchor_map, par_list)
+    else: # use first visible anchor
+        for par in par_list:
+            if par.anchor and par.audio and is_anchor_on_screen(par.anchor):
+                return par
+        # just use first par
+        for par in par_list:
+            if par.audio:
+                return par
+        return None
+
+
+def mark_smil_element(anchor):
+    elem = document.getElementById(anchor)
+    if elem:
+        sel = window.getSelection()
+        sel.selectAllChildren(elem)
+        return bool(sel.rangeCount and sel.toString())
+    return False

src/pyj/read_book/view.pyj

@@ -40,6 +40,7 @@ from read_book.scrollbar import BookScrollbar
 from read_book.search import SearchOverlay
 from read_book.selection_bar import SelectionBar
 from read_book.shortcuts import create_shortcut_map
+from read_book.smil import get_smil_audio_map
 from read_book.timers import Timers
 from read_book.toc import get_current_toc_nodes, update_visible_toc_nodes
 from read_book.touch import set_left_margin_handler, set_right_margin_handler
@@ -328,8 +329,9 @@ class View:
         self.overlay = Overlay(self)
         self.selection_bar = SelectionBar(self)
         self.read_aloud = ReadAloud(self)
+        self.read_audio_ebook = ReadAudioEbook(self)
         self.hints = Hints(self)
-        self.modal_overlays = v'[self.selection_bar, self.read_aloud, self.hints]'
+        self.modal_overlays = v'[self.selection_bar, self.read_aloud, self.hints, self.read_audio_ebook]'
         self.processing_spine_item_display = False
         self.pending_load = None
         self.currently_showing = {'selection': {'empty': True}, 'on_load':v'[]'}
@@ -679,7 +681,7 @@ class View:
         self.iframe.contentWindow.focus()
 
     def start_read_aloud(self, dont_start_talking):
-        if self.is_audio_ebook:
+        if self.book.manifest.has_smil:
             for x in self.modal_overlays:
                 if x is not self.read_audio_ebook:
                     x.hide()
@@ -937,7 +939,6 @@ class View:
         self.book = current_book.book = book
         hl = None
         if not is_redisplay:
-            self.is_audio_ebook = undefined
             if self.read_audio_ebook:
                 self.read_audio_ebook.hide()
                 clear(self.read_audio_ebook.container)
@@ -993,26 +994,6 @@ class View:
                 show_controls_help()
                 sd.set('controls_help_shown_count' + ('_rtl_page_progression' if rtl_page_progression() else ''), c + 1)
 
-        if self.is_audio_ebook is undefined:
-            smil_files = []
-            audio_files = []
-            for filename in book.manifest.files:
-                if filename.endswith(".smil"):
-                    smil_files.append(filename)
-                elif book.manifest.files[filename].mimetype is "audio/mpeg":
-                    audio_files.append(filename)
-            if len(smil_files) > 0:
-                self.is_audio_ebook = True
-                self.read_audio_ebook = ReadAudioEbook(self)
-                self.read_audio_ebook.smil_files = smil_files
-                self.read_audio_ebook.audio_files = audio_files
-                self.read_audio_ebook.audio_maps = {}
-                for smil_file in smil_files:
-                    ui_operations.get_file(
-                        self.book, smil_file, self.read_audio_ebook.parse_smil_file)
-            else:
-                self.is_audio_ebook = False
-
     def preferences_changed(self):
         self.set_margins()
         ui_operations.update_url_state(True)
@@ -1089,8 +1070,6 @@ class View:
             self.loaded_resources = resource_data
             done_callback(resource_data)
         load_resources(self.book, name, self.loaded_resources, cb)
-        if self.is_audio_ebook and self.read_audio_ebook.state != 'HIDDEN':
-            window.setTimeout(self.read_audio_ebook.change_audio_src, 1000) # wait for previous spine to update
 
     def goto_doc_boundary(self, start):
         name = self.book.manifest.spine[0 if start else self.book.manifest.spine.length - 1]
@@ -1396,9 +1375,14 @@ class View:
         self.processing_spine_item_display = True
         self.current_status_message = ''
         self.iframe.style.visibility = 'hidden'
+        fdata = self.book.manifest.files[self.currently_showing.name]
+        smil_map = None
+        if self.book.manifest.has_smil and fdata:
+            smil_map = fdata.smil_map
+            self.currently_showing.smil_audio_map = get_smil_audio_map(smil_map)
         self.iframe_wrapper.send_unencrypted_message('display',
             resource_data=resource_data, book=self.book, name=self.currently_showing.name,
-            initial_position=self.currently_showing.initial_position,
+            initial_position=self.currently_showing.initial_position, smil_map=smil_map or None,
             settings=self.currently_showing.settings, reference_mode_enabled=self.reference_mode_enabled,
             is_titlepage=self.currently_showing.name is self.book.manifest.title_page_name,
             highlights=self.annotations_manager.highlights_for_currently_showing(),

src/pyj/select.pyj

@@ -53,24 +53,6 @@ def first_visible_word():
             if r?:
                 return r
 
-def span_id_at_point(x, y):
-    elements = document.elementsFromPoint(x, y)
-    for element in elements:
-        spans = element.querySelectorAll('span[id]')
-        if len(spans) > 0:
-            return spans[0].id
-
-def id_of_first_visible_span():
-    width = window.innerWidth
-    height = window.innerHeight
-    xdelta = width // 10
-    ydelta = height // 10
-    for y in range(0, height, ydelta):
-        for x in range(0, width, xdelta):
-            span_id = span_id_at_point(x, y)
-            if span_id:
-                return span_id
-
 def empty_range_extents():
     return {
         'start': {'x': 0, 'y': 0, 'height': 0, 'width': 0, 'onscreen': False, 'selected_prev': False},

src/pyj/viewer-main.pyj

@@ -6,7 +6,7 @@ from elementmaker import E
 
 import initialize  # noqa: unused-import
 import traceback
-from ajax import ajax, workaround_qt_bug
+from ajax import absolute_path, ajax, workaround_qt_bug
 from book_list.globals import get_session_data, set_session_data
 from book_list.library_data import library_data
 from book_list.theme import css_for_variables, get_color
@@ -56,6 +56,10 @@ def file_received(name, file_data, proceed, end_type, xhr, ev):
     proceed(result, name, file_data.mimetype, book)
 
 
+def get_url_for_book_file_name(name):
+    return absolute_path('book/' + name)
+
+
 def get_file(book, name, proceed):
     entry = book.manifest.files[name]
     if not entry:
@@ -322,6 +326,7 @@ if window is window.top:
     if TRANSLATIONS_DATA:
         install(TRANSLATIONS_DATA)
     ui_operations.get_file = get_file
+    ui_operations.get_url_for_book_file_name = get_url_for_book_file_name
     ui_operations.get_mathjax_files = get_mathjax_files
     ui_operations.update_url_state = update_url_state
     ui_operations.show_error = show_error