load()
ApiObjects
Flattened
Flattened.chapters
Flattened.paragraphs
Flattened.tables
Flattened.figures
Flattened.rects
Root
File
File.id_
FileMeta
Page
Page.id_
Element
Element.id_
Element.uid
Chapter
Chapter.id_
Paragraph
Paragraph.id_
Paragraph.set_links_backref()
Table
Table.id_
Table.rows
Table.columns
Table.rows_count
Table.columns_count
Cell
Cell.set_links_backref()
Figure
Figure.id_
Figure.set_links_backref()
Rect
Rect.id_
Position
Link
Link.source_chars
Char
Word
Word.text
HorizontalLine
HorizontalLine.text
HorizontalBox
HorizontalBox.text
HorizontalBox.words
![.. uml::
@startuml
skinparam object {
AttributeFontSize 11
}
' increase distance between nodes
skinparam nodesep 50
' default arrow color and arrow font color
!define COL_ARROW_DEFAULT darkblue
skinparam arrowcolor COL_ARROW_DEFAULT
skinparam arrowfontcolor COL_ARROW_DEFAULT
skinparam shadowing false
skinparam package<<Layout>> {
borderColor Transparent
backgroundColor Transparent
fontColor Transparent
stereotypeFontColor Transparent
}
' color of arrows showing inheritance
!define COL_DERIVE red
class Root {
--
**content**: container for elements of type
""Chapter/Table/Figure/Paragraph""
--
**library/API**: All ""Root"" elements are exposed as given in this diagram.
**CLI/YAML/JSON**: ""Root.pages"" contain a list of ""Page"" instances without
""Page.contents"" attribute. Instead, all content goes to ""Root.contents""
and the nested ""Chapter.contents"".
}
class Element {
+ [str] id
+ [str] type
--
**id**: unique in given scope (root or chapter)
**type**: parent class name
**b_root**: set, if the element
is directly located under root
**b_chapter**: set, if the element
is located under a chapter
}
class Position {
+ [float] x0
+ [float] y0
+ [float] x1
+ [float] y1
--
**page**: reference to a ""Page"" instance;
For YAML/JSON output it is serialized to
an attribute holding ""Page.id""
""Position"" is either referenced by a ""Cell""
or by an ""Element"" (never both).
**b_cell**: set, if the Position is referenced
by a ""Cell""
**b_element**: set, if the Position is
referenced by an ""Element""
}
class File {
+ [str] id
+ [str] name
+ [str] path
+ [int] page_count
+ [float] crop_top [points]
+ [float] crop_bottom [points]
+ [float] crop_left [points]
+ [float] crop_right [points]
--
**id**: ""file.<name-as-identifier>""
}
class FileMeta {
+ [str] author
+ [str] title
+ [str] subject
+ [str] creator
+ [str] producer
+ [str] keywords
+ [utc_date] creation_date
+ [utc_date] modified_date
+ [bool] trapped
}
class Page {
+ [str] id
+ [int] number
+ [float] width [points]
+ [float] height [points]
--
**id**: ""page.<1,2,3,n>""
**number**: 1-based
**content**: container for elements of type
""Chapter/Table/Figure/Paragraph""
}
class Link {
+ [int] idx_start
+ [int] idx_end
+ [dict] pos_target
+ [str] libpdf_target
--
**pos_target**: it's a dictionary with the position info
e.g.
page: 3,
x: 300.454
y: 300.454
**libpdf_target**: points either to an ""Element"" or
to a ""Page"". The link is built by concatenating
nested elements separated by '/', e.g.
"" chapter.3/chapter.3.2/table.2""
For case where the pos_target can not be resolved,
the target is set to the target coordinates given as page.<id>/<X>:<Y>
"" page.4/56:789 ""
}
class Paragraph {
--
**id**: ""paragraph.<1,2,3,n>""
A paragraph gets detected by
layout analysis. Characters are
merged into words, words into
lines and lines into paragraphs.
A new word, line or paragraph
is started if the gap changes
significantly.
}
class Chapter {
+ [str] title
+ [str] number
--
**id**: ""chapter.<number>""
**number**: globally unique (e.g. 3.2.4)
**content**: container for elements of type
""Chapter/Table/Figure/Paragraph""
HorizontalBox is None, when the search algorithm
did not find a suitable text box
on the page (aka. ghost chapter).
}
class Cell {
+ [int] row
+ [int] col
--
**row**: 1-based
**col**: 1-based
}
class Table {
**id**: ""table.<1,2,3,n>""
}
class Figure {
+ [str] rel_path
+ [str] caption
--
**id**: ""figure.<1,2,3,n>""
**rel_path**: figures/<image>
}
class Rect {
+ [tuple] non_stroking_color
--
**id**: ""rect.<1,2,3,n>""
**non_stroking_color**: RGB color space,
given as floats between 0 and 1
}
package x <<layout>> {
class HorizontalBox {
+ [float] x0
+ [float] y0
+ [float] x1
+ [float] y1
+ [str] text
--
**text**: plain text in the box
}
class HorizontalLine {
+ [float] x0
+ [float] y0
+ [float] x1
+ [float] y1
+ [str] text
--
**text**: plain text in the line
}
class Word {
+ [float] x0
+ [float] y0
+ [float] x1
+ [float] y1
+ [str] text
--
**text**: plain text in the word
}
class Char {
+ [float] x0
+ [float] y0
+ [float] x1
+ [float] y1
+ [str] text
--
**text**: the character itself
}
}
Char "+chars 1..*" ---left* Word
Word "+words 1..*" ---left* HorizontalLine
HorizontalLine "+lines 1..*" --* HorizontalBox
HorizontalBox "+textbox 1" --* Paragraph
HorizontalBox "+textbox 0..1" --* Cell
HorizontalBox "+textboxes 0..*" --* Figure
HorizontalBox "+textbox 0..1" --* Chapter
HorizontalBox "+textbox 0..1" --* Rect
Paragraph "+b_source 1" *-- "+links *" Link
Figure "+b_source 1" *-- "+links *" Link
Cell "+b_source 1" *-- "+links *" Link
Cell "+cells 1..*" --* "+b_table 1" Table
Cell "+b_cell 1" *-down- "+position 1" Position
Table -[#COL_DERIVE]|> Element
Figure -[#COL_DERIVE]-|> Element
Chapter -[#COL_DERIVE]-|> Element
Paragraph -[#COL_DERIVE]|> Element
Rect -[#COL_DERIVE]|> Element
Chapter "1" *-- "+content *" Element: ordered
Page "1" *-- "+content *" Element: ordered
' all root structures derive from Element
Position "+position 1" --* "+b_element 1" Element
Element "+content *" --* "+b_root 1" Root: ordered
File "+file 1" --* "+b_root 1" Root
Page "+pages 1..*" --* "+b_root 1" Root
Position "+b_positions *" --* "+page 1" Page
FileMeta "+file_meta 1" --* "+b_file 1" File
@enduml](../_images/plantuml-5bb7c06c7cd0021509c933d56fd051c844d7d798.png)