How to Convert HTML to JSON in PowerShell

To convert HTML to JSON in PowerShell, pull the data out of the HTML into objects first, then pipe the objects to ConvertTo-Json. For a clean table, the quickest way is to cast the HTML to [xml]:

$table = [xml](Get-Content -Path .\stores.html -Raw)
$table.table.tr | Select-Object -Skip 1 | ForEach-Object { [pscustomobject]@{ Store = $_.td[0]; City = $_.td[1] } } | ConvertTo-Json

Real web pages are rarely clean enough for that. Below I cover the [xml] route, why it breaks on everyday HTML, and a small function that turns any simple table into JSON.

Everything runs on local files with no extra modules, in PowerShell 7 and Windows PowerShell 5.1.

Convert a clean HTML table with [xml]

My stores.html is a small, well-formed table, like the ones ConvertTo-Html produces. [xml] turns it into objects with tr and td properties:

$table = [xml](Get-Content -Path .\Html\stores.html -Raw)
$rows  = $table.table.tr | Select-Object -Skip 1
$stores = foreach ($tr in $rows) {
    [pscustomobject]@{ Store = $tr.td[0]; City = $tr.td[1]; Sales = [int]$tr.td[2] }
}
$stores | ConvertTo-Json

Output:

[
  {
    "Store": "Downtown",
    "City": "Austin",
    "Sales": 1250
  },
  {
    "Store": "Riverside",
    "City": "Denver",
    "Sales": 980
  }
]

Cast numbers yourself, as with [int] here, or they end up as strings in the JSON. If your table came from ConvertTo-Html, see creating an HTML table in PowerShell.

Why [xml] fails on real web pages

HTML allows things XML doesn’t, like a <br> without a closing tag or a <p> that’s never closed. My report.html has both, and [xml] gives up:

try   { $page = [xml](Get-Content -Path .\Html\report.html -Raw) }
catch { "Failed. " + ($_.Exception.Message -split '\. Error: ')[1] }

Output:

Failed. "The 'br' start tag on line 12 position 18 does not match the end tag of 'td'. Line 12, position 28."

Windows PowerShell 5.1 used to offer Invoke-WebRequest with ParsedHtml for this, but it depends on Internet Explorer and doesn’t exist in PowerShell 7. A small regex function works in both. More on regex is in PowerShell regex.

A ConvertFrom-HtmlTable function

This function finds the rows, uses the <th> cells as property names, strips tags inside cells, decodes entities like &amp; and turns numbers like 1,250 into real numbers:

function ConvertFrom-HtmlTable ([string]$Html) {
    $clean = { param($s) [System.Net.WebUtility]::HtmlDecode(($s -replace '<br\s*/?>', ' ' -replace '<[^>]+>', '').Trim()) }
    $rows = [regex]::Matches($Html, '(?s)<tr[^>]*>(.*?)</tr>') | ForEach-Object { $_.Groups[1].Value }
    $head = [regex]::Matches($rows[0], '(?s)<th[^>]*>(.*?)</th>') | ForEach-Object { & $clean $_.Groups[1].Value }
    foreach ($row in $rows | Select-Object -Skip 1) {
        $cells = @([regex]::Matches($row, '(?s)<td[^>]*>(.*?)</td>') | ForEach-Object { & $clean $_.Groups[1].Value })
        $obj = [ordered]@{}
        for ($i = 0; $i -lt $head.Count; $i++) {
            $v = $cells[$i]
            if ($v -match '^-?[\d,]+$')            { $v = [long]($v -replace ',', '') }
            elseif ($v -match '^-?[\d,]+\.\d+$') { $v = [decimal]($v -replace ',', '') }
            $obj[$head[$i]] = $v
        }
        [pscustomobject]$obj
    }
}

report.html has two tables, so I pick the one with id=”stores” first and pass only that:

. .\ConvertFrom-HtmlTable.ps1
$html  = Get-Content -Path .\Html\report.html -Raw
$table = [regex]::Match($html, '(?s)<table[^>]*id="stores"[^>]*>.*?</table>').Value

ConvertFrom-HtmlTable $table | ConvertTo-Json -Compress

Output:

[{"Store":"Downtown","City":"Austin","Sales":1250},{"Store":"Riverside & Main","City":"Denver","Sales":980},{"Store":"Harbor Point","City":"Seattle","Sales":1105}]

Riverside &amp; Main became Riverside & Main, and Harbor<br>Point became Harbor Point. -Compress puts the JSON on one line. Windows PowerShell 5.1 writes the & as \u0026, which is valid JSON and reads back as &. For the -Compress and -Depth options, see PowerShell ConvertTo-Json.

Regex is fine for simple tables you know. It can’t handle tables nested inside cells or merged cells with colspan, so use a real HTML parser module for those.

Convert an HTML list to a JSON array

Lists are simpler than tables. Grab the <li> items from one <ul> and pass them with -InputObject:

$html  = Get-Content -Path .\Html\report.html -Raw
$list  = [regex]::Match($html, '(?s)<ul[^>]*id="todo"[^>]*>(.*?)</ul>').Groups[1].Value
$items = [regex]::Matches($list, '(?s)<li[^>]*>(.*?)</li>') | ForEach-Object { [System.Net.WebUtility]::HtmlDecode($_.Groups[1].Value) }

ConvertTo-Json -InputObject @($items) -Compress

Output:

["Order shelves","Hire 2 cashiers","Fix FAQ page"]

-InputObject with @() keeps the brackets even when the list has a single item. To work with the array afterward, see reading a JSON file into an array.

Save the JSON to a file

Write the result with Set-Content and read it back to make sure the numbers survived:

. .\ConvertFrom-HtmlTable.ps1
$html  = Get-Content -Path .\Html\report.html -Raw
$table = [regex]::Match($html, '(?s)<table[^>]*id="stores"[^>]*>.*?</table>').Value

ConvertFrom-HtmlTable $table | ConvertTo-Json | Set-Content -Path .\Html\stores.json -Encoding utf8
$back = Get-Content -Path .\Html\stores.json -Raw | ConvertFrom-Json
"Saved $($back.Count) stores, total sales $(($back | Measure-Object -Property Sales -Sum).Sum)"
$back | Format-Table

Output:

Saved 3 stores, total sales 3335

Store            City    Sales
-----            ----    -----
Downtown         Austin   1250
Riverside & Main Denver    980
Harbor Point     Seattle  1105

Sales came back as numbers, so Measure-Object can total them. Writing JSON files is covered in writing JSON to a file.

Frequentl