<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Szymon Wojciechowski</title>
    <description>The latest articles on DEV Community by Szymon Wojciechowski (@simonlab).</description>
    <link>https://dev.to/simonlab</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4250600%2F63238821-a491-46c7-9a09-65ca2d56a585.png</url>
      <title>DEV Community: Szymon Wojciechowski</title>
      <link>https://dev.to/simonlab</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/simonlab"/>
    <language>en</language>
    <item>
      <title>Version the project, not the file pile</title>
      <dc:creator>Szymon Wojciechowski</dc:creator>
      <pubDate>Tue, 17 Nov 2026 21:29:03 +0000</pubDate>
      <link>https://dev.to/simonlab/version-the-project-not-the-file-pile-p7c</link>
      <guid>https://dev.to/simonlab/version-the-project-not-the-file-pile-p7c</guid>
      <description>&lt;p&gt;Keeping a plant project on one revision standard is a coordination problem, not a formatting problem. A typical chemical or process design still moves through a chain of files: a process description, a P&amp;amp;ID, an equipment list, a calculation sheet, a cabinet register, an isometric, a structural drawing. Each of those files is often produced by hand from the one before it. When a change arrives, the question in the meeting is still “which file is current?” rather than “what changed in the project, and what else must follow?”&lt;/p&gt;

&lt;p&gt;This article describes a practical way to version the project itself as structured data, generate the documents from that data, and run checks when a change would make two views of the same object disagree. The examples use a synthetic line in a demonstration unit. They are not results from a client project, and they are not a promise of a public tool or API.&lt;/p&gt;

&lt;h2&gt;
  
  
  The broken chain is a versioning failure
&lt;/h2&gt;

&lt;p&gt;In a file-centered project, revision letters live on documents. One spreadsheet can be in revision C, a drawing still in A, and a permit attachment somewhere between the two. Nothing in the file system knows that those three objects describe the same pipeline, or that a value copied from the offer into the data sheet was rounded twice along the way.&lt;/p&gt;

&lt;p&gt;The loss is not only typographical. A pipeline has one engineering definition: a number, a medium, a nominal size, a material, a pipe class, and the equipment it connects. That same definition is also a load for a supporting structure, a pressure drop for a pump, and a range check for a flowmeter. If those relationships exist only in someone’s memory, a clean-looking document set can still be internally inconsistent.&lt;/p&gt;

&lt;p&gt;A useful standard, then, is not “every PDF has a revision in the title block.” It is: the project has a revision, the documents are generated views of that revision, and a change is recorded as a change to the data, with a list of views that must be issued again.&lt;/p&gt;

&lt;h2&gt;
  
  
  One definition, many views
&lt;/h2&gt;

&lt;p&gt;Treat each object as data first. A line, a vessel, an instrument or a cabinet circuit is defined once. The P&amp;amp;ID sheet, the line register and the isometric are not three independent originals. They are three documents that read the same definition.&lt;/p&gt;

&lt;p&gt;That shift has an immediate effect on version control. You are no longer hoping that three editors applied the same red-line. You are storing one change and regenerating the views. The PDF set still matters, because reviewers and contractors read PDFs. It is just no longer the source of truth. The source of truth is the project data. The PDF follows from it.&lt;/p&gt;

&lt;p&gt;This is the idea behind project as code: the plant project is structured, versioned data, and the documentation is generated from that model. A longer illustrated discussion is on the Simon Lab knowledge page: &lt;a href="https://simon-lab.org/knowledge/project-as-code/" rel="noopener noreferrer"&gt;https://simon-lab.org/knowledge/project-as-code/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  A worked change: one line, several consequences
&lt;/h2&gt;

&lt;p&gt;Take a synthetic example used in public Simon Lab material. A process line, identified in the illustrations as a discharge line, is resized from DN50 to DN65. In a file pile, that change might be marked on the P&amp;amp;ID and missed on the isometric, or updated in the line list and left stale on a pump data sheet.&lt;/p&gt;

&lt;p&gt;If the line is one object, the same change reaches every document that is generated from it: the P&amp;amp;ID sheet, the line register and the isometric. The more useful part is what the model can show next, because other objects read that definition:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;The pressure loss in the discharge line drops.&lt;/li&gt;
&lt;li&gt;The flowmeter that was selected for the previous velocity or range is now out of range and has to be selected again.&lt;/li&gt;
&lt;li&gt;The pump may still be in range, even though a smaller, cheaper model would now do.&lt;/li&gt;
&lt;li&gt;A structure that carries the filled line must be recalculated for a different weight.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;None of those follow-on effects “wait until someone notices,” provided the relationships are in the data and the checks actually run. The important point for a reviewer is not a dramatic saving. It is that the inconsistency is visible in the same revision as the change.&lt;/p&gt;

&lt;p&gt;The illustrations of this example on the public site, including a drawing of a line and a revision log, are illustrations. They are not the file formats of a production platform and should not be copied as if they were an API.&lt;/p&gt;

&lt;h2&gt;
  
  
  Checks after a change, the way tests run after a commit
&lt;/h2&gt;

&lt;p&gt;In software, a change is not finished when the file is saved. Tests run, and a failure is a fact about the current revision. A project defined as data can be treated in a similar way. After each change the model is checked against physics, limits and geometry.&lt;/p&gt;

&lt;p&gt;Examples that belong to this way of working, and that appear in the public demonstration of an anonymised NaOH batch dissolution unit, include:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;A batch program running on an emulated PLC, with dosing throttled near a 60 °C limit.&lt;/li&gt;
&lt;li&gt;Clash detection that reports findings in the 3D assembly, shown as findings, not as proof that the clashes have been removed.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;A failing check is a finding to be resolved by an engineer. It is not an automatic edit of the design, and it is not a certificate that the plant is safe. The check only tells you that this revision violates a stated rule: a limit, a physical relation, or a geometric clearance.&lt;/p&gt;

&lt;p&gt;If you are designing a lightweight review process around this idea, a revision record needs at least:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;The identifier of the object that changed, and the field that changed (for example, nominal size DN50 to DN65).&lt;/li&gt;
&lt;li&gt;The reason, in one or two sentences, so the next reader is not reconstructing intent from a drawing cloud.&lt;/li&gt;
&lt;li&gt;The list of generated documents affected, with their document numbers.&lt;/li&gt;
&lt;li&gt;The list of checks that were run, with pass, fail or not applicable, and any findings.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;That record is what you version. The PDFs are issued from it.&lt;/p&gt;

&lt;h2&gt;
  
  
  What this does not require
&lt;/h2&gt;

&lt;p&gt;You do not need a public software platform to start writing project data in a more disciplined way. A carefully maintained register, with stable identifiers and a rule that drawings are issued from the register, already removes some of the drift. The harder part is encoding the dependencies: that this line’s mass is an input to that beam, or that this instrument’s range is a function of that diameter.&lt;/p&gt;

&lt;p&gt;What you should not do is declare victory because a folder is in Git. A repository full of unmanaged PDFs still has the same consistency problem. Version control of files is necessary and not sufficient. The unit of versioning has to be the project object, not the printout.&lt;/p&gt;

&lt;h2&gt;
  
  
  Limitations
&lt;/h2&gt;

&lt;p&gt;This article uses a demonstration and a synthetic line change. It does not report a measured improvement on a client’s project, and it does not claim a percentage reduction in errors, cost or time. Particle-track visuals in related material are visualisations. Clash results in the demonstration are findings, not resolved clashes.&lt;/p&gt;

&lt;p&gt;Simon Lab develops a computational engineering platform that applies this approach, with layers for specification, process, equipment, cabinets, 3D assembly and site. That platform is in private beta with trusted partners. It is not a public SaaS product, and this article does not offer self-serve access, a public API or an integration that has not been published. Opening the source code is planned later and is not available today. Simon Lab also provides chemical and process engineering design services; those services are a separate matter from access to the platform.&lt;/p&gt;

&lt;p&gt;If you want the fuller argument, with the broken-chain examples and the same line-change worked through in context, it is here: &lt;a href="https://simon-lab.org/knowledge/project-as-code/" rel="noopener noreferrer"&gt;https://simon-lab.org/knowledge/project-as-code/&lt;/a&gt;&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Disclosure: this article was drafted with the help of AI and checked by the author, a process engineer with 15 years of design experience.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>engineering</category>
      <category>versioncontrol</category>
      <category>documentation</category>
      <category>data</category>
    </item>
  </channel>
</rss>
