Excel spreadsheet crawler and table parser for data extraction and querying.
- Command-line interface
- Recursive Excel file discovery
- Sub-tabular data extraction (logical tables)
- SQLite and PostgreSQL database interfaces
- CLI query tool
- Summary data metrics
To install eparse, you can use pip and the latest version on PyPI:
$ pip install eparse
Or you can clone this repo and install from source, as the latest version will not always by on PyPI:
$ git clone https://github.com/ChrisPappalardo/eparse.git $ cd eparse $ pip install .
Using eparse in another project? You can also add either a PyPI version
or the latest source to your requirements.txt
file as follows:
eparse # latest pypi version eparse==0.8.0 # sepcific pypi version eparse @ git+https://github.com/ChrisPappalardo/eparse.git # latest source
If you plan to use the postgres interface, you also need to install
the postgres package psycopg2
. Instructions can be found
here.
This package is optional, and you can use the other interfaces
such as the SQLite3
interface without having to install
psycopg2
.
The easiest way to install the psycopg2
package for your
particular environment may be to install the pre-compiled
binary driver as follows:
$ pip install psycopg2-binary
If you see an error while trying to use a postgres endpoint such
as postgres://user:pass@host:port/my_db
that mentions the
postgres driver is missing, then you know you haven't properly
installed (and compiled) psycopg2
.
eparse can be used as either a python library or from the command-line.
You can view supported CLI commands and usage with --help
as follows:
$ eparse --help Usage: eparse [OPTIONS] COMMAND [ARGS]... excel parser Options: -i, --input TEXT input source -o, --output TEXT output destination -f, --file TEXT file(s) or dir(s) to target -d, --debug use debug mode -l, --loose find tables loosely -r, --recursive find files recursively -t, --truncate truncate dataframe output -v, --verbose increase output verbosity --help Show this message and exit. Commands: migrate migrate eparse table parse parse table(s) found in sheet for target(s) query query eparse output scan scan for excel files in target
You can also use eparse from python like so:
from eparse.core import get_df_from_file print([table for table in get_df_from_file('myfile.xlsx')]) 102 Date Principal Repayment Date Principal Repayment 103 44834 700757.679004 44926 430013.148303 104 44926 71957.776108 45016 100576.127808 105 45016 147578.19262 45107 898008.340095 106 45107 32801.363072 45199 841656.13896 ...
To scan one or more directories for Excel files with descriptive
information, you can use the scan
command like so:
$ eparse -v -f <path_to_files> scan
Increase the verbosity with additional flags, such as -vvv
, for
more descriptive information about the file(s), including sheet names.
Excel files can be parsed as follows:
$ eparse -v -f <path_to_files> parse
This mode will list each table found in each Excel file to the command-line. This mode is useful for initial discovery for parseable data.
eparse uses a simple algorithm for identifying tables. Table "corners"
are identified as cells that contain empty cells above and to the left
(or sheet boundaries). A densely or sparsely populated 2x2+ table must
follow in order for data to be extracted in relation to that cell.
eparse will automatically adjust for rowspan labels and empty table
corners and the dense vs. sparse criterion can be controlled with
the --loose
flag. eparse can also tolerate a user-specified number
of NA row and column cells and still consider the table to be unbroken
with the --nacount
arg.
eparse was written to accomodate various types of output formats and
endpoints, including null:///
, stdout:///
, sqlite3:///db_name
,
and postgres://user:password@host:port/db_name
.
This mode is useful for validating files and generating descriptive info, and is the default. The command above with -v is an example of this mode, which lists out the tables found.
This mode is good for viewing data extracted from Excel files in the console. For example, you could view all tables found in Sheet1 with the following command:
$ eparse -f <path_to_files> -o stdout:/// parse -s "Sheet1"
eparse uses pandas
to handle table data. You can view larger tables without truncation
using the -t
flag as follows:
$ eparse -t -f <path_to_files> -o stdout:/// parse -s "Sheet1"
Data in table format is useful for human viewing, but a serialized
form is better for data interfacing. Serialize your output with
the -z
flag as follows:
$ eparse -t -f <path_to_files> -o stdout:/// parse -z
Each cell of extracted table data is serialized as follows:
- row - 0-indexed table row number
- column - 0-indexed table column number
- value - the value of the cell as a
str
- type - the implied python
type
of the data found - c_header - the column header
- r_header - the row header
- excel_RC - the RC reference from the spreadsheet (e.g. B10)
- sheet - the name of the sheet
- f_name - the name of the file
eparse uses the peewee
package for ORM and database integration. The
interfaces module contains an
ExcelParse
model that provides data persistence and a common
interface.
To create a SQLite3 database with your parsed Excel data, use the following command:
$ mkdir .files $ eparse -f <path_to_files> -o sqlite3:/// parse -z
This command will automatically generate a unique database filename
using the uuid
python package in the .files/
sub-directory
of the working directory. You may need to create this directory
before running this command, as shown.
You can also specify a path and filename of your choosing, like so:
$ mkdir .files $ eparse -f <path_to_files> -o sqlite3:///path/filename.db parse -z
eparse also supports postgresql integrations. As mentioned above,
you will need psycopg2
installed for postgresql integrations
to work. The eparse BaseDatabaseInterface
abstracts the
implementation details, so you would use this interface the same
way you use the others, with the exception of the endpoint.
To use a postgresql
database as the source and/or destination
of your data, you would supply an --input
and/or --output
endpoint to the tool as follows:
$ eparse -o postgres://user:password@host:port/db_name ...
Where details like user
, host
, port
are provided to
you by your db administrator. eparse will create the necessary
table(s) and indexes for you when inserting data into the database.
Once you have stored parsed data, you can begin to query it using the
peewee
ORM. This can be done with the tool or directly with
the database.
For example, query distinct column header names from a generated
SQLite3
database as follows:
$ eparse -i sqlite3:///.files/<db_file> -o stdout:/// query -m get_c_header c_header Total Rows Data Types Distinct Values 0 ABC-col 150 2 76 1 DEF-col 3981 3 15 2 GHI-col 20 1 2 .. ... ... ... ...
This command will give descriptive information of each distinct c_header found, including total rows, unique data types, and distinct values.
You can also get raw un-truncated data as follows:
$ eparse -t -i sqlite3:///.files/<db_file> -o stdout:/// query
Filtering data on content is easy. Use the --filter
option as
follows:
$ eparse -i sqlite3:///.files/<db_file> -o stdout:/// query --filter f_name "somefile.xlsx"
The above command will filter all rows from an Excel file named
somefile.xlsx. You can use any of the following django
-style
filters:
__eq
equals X__lt
less than X__lte
less than or equal to X__gt
greater than X__gte
greater than or equal to X__ne
not equal to X__in
X is in__is
is X__like
like expression, such as%somestr%
, case sensitive__ilike
like expression, such as%somestr%
, case insensitive__regexp
regular expression matching such as^.*?foo.*?$
Filters are applied to the ORM fields like so:
--filter row__gte 4
all extracted table rows >= 5--filter f_name__ilike "%foo%"
all data from filenames with foo--filter value__ne 100
all data with values other than 100
Queried data can even be stored into a new database for creating curated data subsets, as follows:
$ eparse -i sqlite3:///.files/<db_file> \ -o sqlite3:///.files/<subq_db_file> \ query --filter f_name "somefile.xlsx"
Since database files the tool generates when using sqlite3:/// are
SQLite
native, you can also use SQLite database client tools
and execute raw SQL like so:
$ sudo apt-get install -y sqlite3-tools $ sqlite3 .files/<db_file> SQLite version 3.37.2 2022-01-06 13:25:41 Enter ".help" for usage hints. sqlite> .schema CREATE TABLE IF NOT EXISTS "excelparse" ("id" INTEGER NOT NULL PRIMARY KEY, "row" INTEGER NOT NULL, "column" INTEGER NOT NULL, "value" VARCHAR(255) NOT NULL, "type" VARCHAR(255) NOT NULL, "c_header" VARCHAR(255) NOT NULL, "r_header" VARCHAR(255) NOT NULL, "excel_RC" VARCHAR(255) NOT NULL, "name" VARCHAR(255) NOT NULL, "sheet" VARCHAR(255) NOT NULL, "f_name" VARCHAR(255) NOT NULL); sqlite> .header on sqlite> SELECT * FROM excelparse LIMIT 1; id|row|column|value|type|c_header|r_header|excel_RC|name|sheet|f_name 1|0|0|ABC|<class 'str'>|SomeCol|SomeRow|B2|MyTable|Sheet1|myfile.xlsm
eparse wouldn't be a solid tool without the ability to migrate your eparse databases for future code changes. You can apply migrations that ship with future versions of eparse as follows:
$ eparse -i sqlite3:///.files/<db_file> migrate -m <migration> applied <migration>
It is up to you to determine the migrations you need based on the eparse version you are upgrading from and to. Migrations can be found in eparse/migrations.py
If you would like to use eparse to partition xls[x] files alongside unstructured, you can do so with our contributed partition and partition_xlsx modules. Simply import the partition function from eparse.contrib.unstructured.partition and use it instead of partition from unstructured.partition.auto like so:
from eparse.contrib.unstructured.partition import partition elements = partition(filename='some_file.xlsx', eparse_mode='...')
Valid eparse_mode settings are available in eparse.contrib.unstructured.xlsx._eparse_modes.
As an open-source project, contributions are always welcome. Please see Contributing for more information.
eparse is licensed under the MIT License. See the LICENSE file for more details.
Thanks for your support of eparse. Feel free to contact me at [email protected] or connect with me on Github.