Skip to content

Simple wrapper of tabula-java: extract table from PDF into pandas DataFrame

License

Notifications You must be signed in to change notification settings

Gallaecio/tabula-py

 
 

Repository files navigation

tabula-py

Build Status PyPI version Patreon

tabula-py is a simple Python wrapper of tabula-java, which can read table of PDF. You can read tables from PDF and convert into pandas's DataFrame. tabula-py also enables you to convert a PDF file into CSV/TSV/JSON file.

Requirements

  • Java
    • Confirmed working with Java 7, 8
  • pandas
  • urllib3
  • distro

OS

I confirmed working on macOS and Ubuntu. But some people confirm it works on Windows 10. See also following the setting procedure.

Usage

Install

pip install tabula-py

If you want to become a contributor, you can install dependency for development of tabula-py as follows:

pip install -r requirements.txt -c constraints.txt

Example

tabula-py enables you to extract table from PDF into DataFrame and JSON. It also can extract tables from PDF and save file as CSV, TSV or JSON.

import tabula

# Read pdf into DataFrame
df = tabula.read_pdf("test.pdf", options)

# Read remote pdf into DataFrame
df2 = tabula.read_pdf("https://github.com/tabulapdf/tabula-java/raw/master/src/test/resources/technology/tabula/arabic.pdf")

# convert PDF into CSV
tabula.convert_into("test.pdf", "output.csv", output_format="csv")

# convert all PDFs in a directory
tabula.convert_into_by_batch("input_directory", output_format='csv')

See example notebook

Get tabula-py working (Windows 10)

This instruction is originally written by @lahoffm. Thanks!

  • If you don't have it already, install Java
  • Try to run example code (replace the appropriate PDF file name).
  • If there's a FileNotFoundError when it calls read_pdf(), and when you type java on command line it says 'java' is not recognized as an internal or external command, operable program or batch file, you should set PATH environment variable to point to the Java directory.
  • Find the main Java folder like jre... or jdk.... On Windows 10 it was under C:\Program Files\Java
  • On Windows 10: Control Panel -> System and Security -> System -> Advanced System Settings -> Environment Variables -> Select PATH --> Edit
  • Add the bin folder like C:\Program Files\Java\jre1.8.0_144\bin, hit OK a bunch of times.
  • On command line, java should now print a list of options, and tabula.read_pdf() should run.

Options

  • pages (str, int, list of int, optional)
    • An optional values specifying pages to extract from. It allows str, int, list of int.
    • Example: 1, '1-2,3', 'all' or [1,2]. Default is 1
  • guess (bool, optional):
    • Guess the portion of the page to analyze per page. Default True
  • area (list of float, optional):
    • Portion of the page to analyze(top,left,bottom,right).
    • Example: [269.875, 12.75, 790.5, 561] or [[12.1,20.5,30.1,50.2],[1.0,3.2,10.5,40.2]]. Default is entire page
  • relative_area (bool, optional):
    • If all area values are between 0-100 (inclusive) and preceded by '%', input will be taken as % of actual height or width of the page. Default False.
  • lattice (bool, optional):
    • [spreadsheet option is deprecated] Force PDF to be extracted using lattice-mode extraction (if there are ruling lines separating each cell, as in a PDF of an Excel spreadsheet).
  • stream (bool, optional):
    • [nospreadsheet option is deprecated] Force PDF to be extracted using stream-mode extraction (if there are no ruling lines separating each cell, as in a PDF of an Excel spreadsheet)
  • password (bool, optional):
    • Password to decrypt document. Default is empty
  • silent (bool, optional):
    • Suppress all stderr output.
  • columns (list, optional):
    • X coordinates of column boundaries.
    • Example: [10.1, 20.2, 30.3]
  • output_format (str, optional):
    • Format for output file or extracted object.
    • For read_pdf(): json, dataframe
    • For convert_into(): csv, tsv, json
  • output_path (str, optional):
    • Output file path. File format of it is depends on format.
    • Same as --outfile option of tabula-java.
  • java_options (list, optional):
    • Set java options like -Xmx256m.
  • pandas_options (dict, optional):
    • Set pandas options like {'header': None}.
  • multiple_tables (bool, optional):
    • (Experimental) Extract multiple tables.
    • This option uses JSON as an intermediate format, so if tabula-java output format will change, this option doesn't work.

FAQ

tabula-py does not work

There are several possible reasons, but tabula-py is just a wrapper of tabula-java, make sure you've installed Java and you can use java command on your terminal. Many issue reporters forget to set PATH for java command.

You can check whether tabula-py can call java from Python process with tabula.environment_info() function.

I can't from tabula import read_pdf

If you've installed tabula, it will be conflict the namespace. You should install tabula-py after removing tabula.

pip uninstall tabula
pip install tabula-py

The result is different from tabula-java. Or, stream option seems not to work appropreately

tabula-py set guess option True by default, for beginners. It is known to make a conflict between stream option. If you feel something strange with your result, please set guess=False.

Can I use option xxx?

Yes. You can use options argument as following. The format is same as cli of tabula-java.

read_pdf(file_path, options="--columns 10.1,20.2,30.3")

How can I ignore useless area?

In short, you can extract with area and spreadsheet option.

In [4]: tabula.read_pdf('./table.pdf', spreadsheet=True, area=(337.29, 226.49, 472.85, 384.91))
Picked up JAVA_TOOL_OPTIONS: -Dfile.encoding=UTF-8
Out[4]:
  Unnamed: 0 Col2 Col3 Col4 Col5
0          A    B   12    R    G
1        NaN    R    T   23    H
2          B    B   33    R    A
3          C    T   99    E    M
4          D    I   12   34    M
5          E    I    I    W   90
6        NaN    1    2    W    h
7        NaN    4    3    E    H
8          F    E   E4    R    4

How to use area option

According to tabula-java wiki, there is a explain how to specify the area: https://github.com/tabulapdf/tabula-java/wiki/Using-the-command-line-tabula-extractor-tool#grab-coordinates-of-the-table-you-want

For example, using macOS's preview, I got area information of this PDF:

image

java -jar ./target/tabula-1.0.1-jar-with-dependencies.jar -p all -a $y1,$x1,$y2,$x2 -o $csvfile $filename

given

Note the left, top, height, and width parameters and calculate the following:

y1 = top
x1 = left
y2 = top + height
x2 = left + width

I confirmed with tabula-java:

java -jar ./tabula/tabula-1.0.1-jar-with-dependencies.jar -a "337.29,226.49,472.85,384.91" table.pdf

Without -r(same as --spreadsheet) option, it does not work properly.

I faced ParserError: Error tokenizing data. C error. How can I extract multiple tables?

This error occurs pandas trys to extract multiple tables with different column size at once. Use multiple_tables option, then you can avoid this error.

I want to prevent tabula-py from stealing focus on every call on my mac

Set java_options=["-Djava.awt.headless=true"]. kudos @jakekara

I got ? character with result on Windows. How can I avoid it?

If the encoding of PDF is UTF-8, you should set chcp 65001 on your terminal before launching a Python process.

chcp 65001

Then you can extract UTF-8 PDF with java_options="-Dfile.encoding=UTF8" option. This option will be added with encoding='utf-8' option, which is also set by default.

# This is an example for java_options is set explicitly
df = read_pdf(file_path, java_options="-Dfile.encoding=UTF8")

Replace 65001 and UTF-8 appropriately, if the file encoding isn't UTF-8.

I can't extract file/directory name with space on Windows

You should escape file/directory name yourself.

Contributing

Interested in helping out? I'd love to have your help!

You can help by:

  • Reporting a bug.
  • Adding or editing documentation.
  • Contributing code via a Pull Request.
  • Write a blog post or spreading the word about tabula-py to people who might be able to benefit from using it.

Contributors

Another support

You can also support our continued work on tabula-py with a donation on Patreon.

About

Simple wrapper of tabula-java: extract table from PDF into pandas DataFrame

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 100.0%