A Day at the Theater and the Cinema: Exploring Shanghai’s Entertainment World, 1907–1966

On 10 October 1931, the advertised daytime program at Shanghai’s Sincere amusement garden (先施乐园) brought together several kinds of entertainment. Its six recorded entries included two Beijing-opera pieces, a film, two works described as xinju (“new drama”), and an untitled attraction.

The film was 《血海奇士》. Search for that title in the newly available Shanghai Entertainment dashboard, select the relevant result, and the rest of its programme appears. A film title becomes an entry point into a broader setting: a venue, a date, a newspaper advertisement, and the other entertainments offered alongside it. This small example illustrates what the interface makes possible. It invites users to move between individual records and larger patterns, exploring the variety of Shanghai’s entertainment culture across six decades.

From newspaper advertisements to a research database

Shanghai Entertainment provides an interactive way into a corpus of 139,655 program entries covering the years 1907–1966. Its materials concern cinema, opera, spoken drama, storytelling, comedy, music, acrobatics, magic, and many other advertised attractions. The resource rests on a substantial collective undertaking. Jiang Jin (蔣進, ECNU) directed the collection and compilation of entertainment advertisements published in four major Shanghai newspapers: Shenbao (申報), Xinwenbao (新聞報), Xinwen ribao (新聞日報), and Jiefang ribao (解放日報). The resulting materials appeared in 2015 as the four-volume 二十世纪上海报刊娱乐版广告资料长编.

The project was originally conceived as a publication. Information was assembled in Word tables, a format suited to preparing the volumes but difficult to use for systematic searches and comparisons. During the final editing phase,  I began discussing with Jiang Jin how to transform this material into a database. Jean-Pierre Dedieu brought his extensive experience with FileMaker to that work, and students participated in transferring and structuring the information.

The online dashboard is a new stage in this history. It makes an existing scholarly database available through a browser, offering a focused set of tools for searching, comparing, and following connections within the records.

Explore Shanghai Entertainment, 1907–1966

What does a “program entry” represent?

Understanding the database begins with a distinction between a show and a performed item. A show is a grouping of entries belonging to a program. It may contain a single film or theatrical work, but it may also bring together several pieces or attractions. A performed item is one of those components: a film, an opera piece, an act, or another advertised entertainment.

The figure of 139,655 therefore refers to program entries, rather than to 139,655 distinct works. A title can recur on different dates or at different venues. The same story may also appear under several genre labels. The database links these items to their program context, with fields for dates, venues, advertised prices, sources, genre designations, daytime or evening scheduling, promotional wording, and named performers where recorded. These relationships allow users to retrieve a whole program from one of its components. The dashboard’s live counters report matching records in the current export. They can differ from the published corpus total while the export is being reconciled with the source database.

One program, several forms of entertainment

The Sincere example shows why preserving program context matters.

Searching for 《血海奇士》 retrieves a film entry. Opening its programme for 10 October 1931 reveals two entries classified as Beijing opera—《狸猫》 and 《西游记》—alongside the film, two xinju entries—《云外飘香》 and 《马永贞》—and an untitled attraction. The recorded source is Xinwenbao for that date. The example establishes what the database records as an advertised programme; it does not establish how an individual visitor experienced the venue or whether every announced performance took place.

Even with that distinction, the combination is suggestive. It encourages us to examine how entertainment forms were assembled within particular venues. How did programs vary between amusement halls and specialized theaters? Which combinations recurred? Did the balance change over time? A database organized only as a list of titles would make those questions harder to pursue. Reconstructing the program restores an important layer of context.

Illustration: Searching for 《血海奇士》 and selecting the Xinwenbao record reveals six advertised daytime entries at 先施乐园 on 10 October 1931. The screenshot shows the search result and its reconstructed programme.

Following stories, names, and places

There are several ways to begin exploring.

A familiar title offers one route. A search for 杨乃武, for example, returns entries under categories including storytelling, Shanghai opera, spoken drama, and cinema. Such results provide leads for investigating how stories moved between forms of performance. They also raise questions that require further reading: does a shared title refer to the same adaptation, a different episode, or a substantially different work? A performer’s name offers another route. Users can inspect associated entries and their distribution over time. These results are records of named appearances in the corpus, not complete biographies or comprehensive career histories. Stage names, spelling variants, and uneven reporting require attention.

A venue can become the focus of an investigation in its own right. Selecting a theater or amusement hall allows users to examine the kinds of entertainment documented there, search its programs, and compare different periods. Venue names sometimes appear in abbreviated or variant forms, so identifying an establishment across the whole corpus may require additional work.

The interface also includes a performer–venue network. This can help users identify recurring associations worth investigating. A shared venue does not, by itself, demonstrate that performers belonged to the same troupe or appeared together. For students, these routes offer manageable starting points. One title, one venue, or a short period can generate a focused exercise in asking historical questions, tracing references, and distinguishing what a record establishes from what it merely suggests.

Reading patterns and their limits

The dashboard provides charts of entries over time, genre distributions, and venue activity. These views help users locate patterns and decide where to look more closely. Their interpretation depends on understanding how the corpus was assembled.

The collection is based on sampling, including the first day of each month and additional festivals and commemorative dates. It is not a continuous census of Shanghai’s entertainment offerings. The newspapers represented also change across the period: the earlier decades draw on Shenbao and Xinwenbao, followed by Xinwen ribao and, for 1960–1966, Jiefang ribao.

The sources presented practical difficulties. The compilation relied on microfilm and reprints, with problems of legibility and missing pages. Advertisements themselves could contain errors or ambiguous abbreviations. Some fields remain incomplete. Consequently, a rise or fall in the number of recorded entries may reflect changes in sampling, surviving newspaper coverage, or advertising practices as well as changes in entertainment activity. A frequently advertised venue cannot automatically be described as the most popular. An advertised offering tells us neither how many people attended nor what they thought of it.

The genre labels also deserve close reading. They preserve the vocabulary recorded in the sources, including compound designations, promotional expressions, and variant terms. For some kinds of quantitative comparison, researchers will want to construct broader classifications. For other questions, those historical distinctions are part of the evidence.

Finally, the database captures selected textual information. It does not reproduce the visual experience of the newspaper page: the size and position of an advertisement, its typography, or its relationship to surrounding material. Source references provide a route for further investigation, but the dashboard does not include scans of the original advertisements. These features make the database particularly useful for exploration and question formation, followed by closer examination of the relevant evidence.

Entertainment took place in a changing city. The locations of theaters, cinemas, amusement halls, and other facilities offer another way to connect these records to Shanghai’s urban history. The repository includes supplementary files concerning facility addresses, historical names, and GIS data. These provide material for further spatial research, while Virtual Shanghai offers a wider setting in which to investigate the city’s historical geography. The current dashboard does not include an integrated map. Linking the supplementary spatial materials to its records requires checking identifiers, name variants, and coordinate information. This is a promising extension for researchers interested in the relationships between entertainment, commercial districts, and urban change.

What LLM assistance changed

There is a second story behind this release: the process of building the interface.

Using a large language model (Claude Sonnet 5), I was able to create the interactive dashboard in approximately 30–40 minutes, working from the existing database. That time refers to developing an interface around structured material. The collection, transcription, organization, and scholarly understanding of that material grew out of years of work. The speed matters because it changes what becomes practical. Researchers can try different ways of exploring a corpus while their questions are still developing. They can give colleagues and students access to selections and relationships that previously required familiarity with the original database software. A chart or example can lead directly to records that others can examine for themselves.

The experience also underlined the importance of checking the results. Reviewing the export and the application revealed date errors, discrepancies between published totals and export counts, and ambiguous identifiers affecting some performer links. The confirmed dates have been corrected. Performer links that cannot be assigned to a single item are omitted from the relevant views pending recovery of their original relationships; the underlying records remain preserved. Rapid development makes verification more valuable. The ability to produce an interface quickly creates time and opportunity to examine its assumptions, test its queries, and improve its documentation. For historical research, the most useful outcome is an easier passage between a substantial body of evidence, a specific question, and the records needed to investigate it.

Explore, reuse, and contribute

Shanghai Entertainment is available online without installing the original database software. The historical records are in Chinese, while the interface and guide are in English. Browse results can be downloaded as CSV files, up to the 2,000 records displayed in a selection.

The database and supplementary datasets are licensed under Creative Commons Attribution–NonCommercial 4.0 International. They may be shared and adapted for non-commercial purposes with appropriate attribution, a link to the licence, and an indication of changes. The software is available under the MIT licence.

Open the dashboardVisit the repository and supplementary datasets

I welcome corrections, suggestions, and accounts of how colleagues and students use the resource. For questions about in-depth research, the original FileMaker database, or commercial-use permission, please contact enpmuc[at]gmail.com.