Parsel: Parse PDFs, Office Documents, and Images in PHP

Parsel parses PDFs, Word documents, spreadsheets, presentations, and images through a single fluent PHP API. Built by 普什帕克·查杰德 , it wraps the liteparse lit CLI and performs all work locally, so files are never sent to an external service. It requires PHP 8.4, and Parsel includes a command to install the lit binary.

# A single API across file types

The same fluent API supports all formats. Call text() for plain text or parse() for a structured Document 目的：

使用

Shipfastlabs\Parsel
;
文本
=

Parsel
::
文件
（
'contract.pdf'
）
->
文本
（）；
$document
=

Parsel
::
文件
（
'amendment.docx'
）
->
解析
（）；

返回的 Document exposes the full text, document metadata, a pageCount() ，和一个 pages collection. Each page carries its number , width , height , text , and a list of positioned items 。

# Positioned text with coordinates

Every text item includes its location and font information, which is useful for reading tables, finding signature blocks, or mapping clauses back to their position on the page. Each item has text , x , y , width , height , fontName , fontSize ，和一个 confidence score:

$document
=

Parsel
::
文件
（
'contract.pdf'
）
->
解析
（）；
foreach
（文档）
->
页面
作为
$page) {

foreach
（每页）
->
项目
作为
$item) {

回声

"{
$item
->
文本
} @ ({
$item
->
十
}, {
$item
->
和
}）
\n
“
;
}
}

# Page selection and streaming

You can limit work to specific pages with page() , pages() , pageRange() ， 或者 maxPages() 。 page() takes a single page number, pages() accepts a list of numbers or range strings, and pageRange() takes a start and end:

// Just the cover page
摘要
=

Parsel
::
文件
（
'contract.pdf'
）
->
页
（
1
）
->
文本
（）；
// A mix of individual pages and a range
$选定
=

Parsel
::
文件
（
'contract.pdf'
）
->
页面
（
'1-3'
,
12
）
->
解析
（）；
// A continuous range of clauses
$body
=

Parsel
::
文件
（
'contract.pdf'
）
->
pageRange
（
2
,
6
）
->
解析
（）；

These calls are additive, so you can combine them before parsing, and maxPages() caps the total number of pages processed:

$document
=

Parsel
::
文件
（
'contract.pdf'
）

->
pageRange
（
1
,
5
）

->
页
（
12
）

->
maxPages
（
20
）

->
解析
（）；

For large files, lazyPages() processes one page at a time to keep memory use flat:

foreach
（
Parsel
::
文件
（
'contract.pdf'
）
->
lazyPages
（）
作为
$page) {

// handle one page at a time
}

# OCR for scanned documents

OCR is off by default. Turn it on with withOcr() and pass named arguments for the language, tessdata path, an OCR server URL, and worker count:

文本
=

Parsel
::
文件
（
'signed-contract.png'
）

->
withOcr
（

语言
:
'eng'
,

tessdataPath
:
'/usr/share/tessdata'
,

serverUrl
:
'http://localhost:8828/ocr'
,

workers
:
8
,
）

->
文本
（）；

# Rendering page previews

Turn pages into image files with screenshots() , passing an output directory:

$screenshots
=

Parsel
::
文件
（
'contract.pdf'
）
->
截图
（
存储路径
（
'previews'
））；

# Passwords and rendering options

Encrypted files open with withPassword() , and a few chainable methods adjust how lit renders a document before parsing. withDpi() raises the render resolution, which sharpens both OCR and screenshots; preserveSmallText() keeps fine print such as footnotes from being dropped; and withTimeout() sets a per-file time limit so a large document cannot stall a request:

$document
=

Parsel
::
文件
（
'contract.pdf'
）

->
withPassword
（
'hunter2'
）

->
withDpi
（
300
）

->
preserveSmallText
（）

->
withTimeout
（
120
）

->
解析
（）；

# Testing without the binary

Parsel ships a fake runner so tests don't have to shell out to the real lit binary. You map command fragments to canned output and assert on the commands that were recorded:

$fake
=

Parsel
::
伪造的
（[

'--format json'

=>

file_get_contents
（
__你__
。
'/fixtures/contract.json'
),
]);
$document
=

Parsel
::
文件
（
'contract.pdf'
）
->
解析
（）；
预计
($fake
->
recordedCommands
（）[
0
]）
->
toContain
（
'--format'
,
‘json’
（英文）：

# 安装

Install the package with Composer:

作曲家

要求

shipfastlabs/parsel

Then install the lit 二进制：

vendor/bin/parsel-install-lit

For Office documents and images, you can pull in the additional system dependencies:

vendor/bin/parsel-install-lit

--with-system-dependencies

If you'd rather manage lit yourself, for example, in a CI image that already has it, the binary is also available through common package managers (npm, pnpm, bun, pip, cargo) and can be installed independently.

You can view the source and full documentation on GitHub 。